Shayna Korol | Vox

The case for AI realism

2026-04-14T16:44:20-04:00

Director Charlie Tyrell, left, and producer Daniel Kwan at a screening of Focus Features’ The AI Doc: or How I Became An Apocaloptimist on March 23, 2026 in Los Angeles, California. | Eric Charbonneau/Focus Features via Getty Images

In 1964, science fiction writer Arthur C. Clarke predicted that computers would overtake human evolution.“Present-day electronic brains are complete morons, but this will not be true in another generation,” he told the BBC. “They will start to think, and eventually, they will completely out-think their makers.”

Daniel Roher opens his new documentary The AI Doc: Or How I Became an Apocaloptimist (2026) with this cheerful prophecy. And in the hundred-some minutes that follow, he tries to make sense of a technology that, by his own admission, he does not understand — and a world that is rapidly being changed by it. Explaining that he conceives of AI as a “magic box floating in space,” he enlists the help of experts to provide him with a crash course in what, exactly, AI is.

Roher’s real concern, however, isn’t so much about the workings of AI — though some of his subjects do attempt to explain them for him — but whether it might displace us, as Clarke’s prediction suggests it will.

While making the film, Roher learns that his wife Caroline is pregnant with their first child. He tracks his wife’s pregnancy and the birth of his son in parallel with the advent of AI. It’s a smart choice that builds on a fear all parents share: What sort of world are we making for our children? And behind that question is another, vibrating in anxious silence: What happens after our offspring replace us? This twinned existential angst drives his efforts to hear from the doomers, the techno-optimists, and the in-between “apocaloptimists” whose ranks he ultimately joins.

The AI Doc, as its sweeping title suggests, wants to shape and lead the narrative around AI. It’s certainly set up to do that — Roher is fresh off an Oscar win for his documentary Navalny, and the film opened in nearly 800 theaters, which counts as wide-release for a nonfiction title. The final product is indicative of the ways that public attitudes around AI are in massive flux. Roher hopes to reach people of my grandmother’s generation who conflate AI with smartphones and spellcheck, as well as people who don’t seem to care whether a video was AI-generated.

But I think that this documentary has come too late to steer the conversation, something the film itself acknowledges. For all its transformative potential, AI isn’t actually unique among emerging technologies yet — it has not been cataclysmic or ushered in a golden age of prosperity — but Roher and many of those he interviews tend to treat it as a radical break with all that has come before. As a result, they tend to fixate on the binary extremes of doom or salvation. It’s an approach that reinforces our own helplessness in the face of AI-driven change, while also muddying our understanding of what we might yet be able to do as we seek to adapt, mitigate harm, and shape the world that AI could otherwise truly start remaking.

For good and for ill

Roher, contemplating his child’s future, opts to hear the bad news first. Tristan Harris, the cofounder of the Center for Humane Technology, doesn’t mince words: “I know people who work on AI risk who don’t expect their children to make it to high school.”

Many of the film’s other interviewees are similarly gloomy. Geoffrey Hinton, the “godfather of AI,” for example, argues that as AI becomes smarter, it will become better at manipulating humanity. But no one is more pessimistic than Eliezer Yudkowsky, the well-known AI doomer and co-author of the controversial book If Anyone Builds It, Everyone Dies. As the title suggests, Yudkowsky believes that superintelligent AI would wipe out humanity — a position that he stands by and lays out for Roher.

Turning his back on these storm clouds — and taking the advice of his wife, Caroline, who tells him that he needs to find hope for the future — Roher tunes into the chorus of AI optimists. They tell him, variously, that there are more potential benefits than downsides to AI; that technology has made the world better in every way; that this will be the tool that helps us solve all our greatest problems. Not to mention: AI will bring the best health care on the planet to the poorest people on Earth, extend our healthspan by decades, and enable us to live in a postscarcity utopia free of drudgery. Oh, and: We will become an interplanetary species, all thanks to AI.

These promises initially reassure Roher, perhaps because he seems easily led by whomever he’s spoken to most recently. It is Harris who ultimately convinces him that we can’t separate the promise of AI from the peril it presents. The conclusions that result will be obvious to anyone who’s thought about these issues for more than a moment or two: If AI automates work, for example, how will people make a living?

It doesn’t help that many of the most invested players reflect on these questions superficially, if at all. OpenAI CEO Sam Altman tells Roher that he’s worried about how authoritarian governments will use AI — a claim that is followed in the film by a cut to images of Altman posing with authoritarian leaders. Other tech CEOs fall back on PR pleasantries in response to the filmmaker’s questions, and Roher too often goes easy on them, never diving deeper when they admit that even they aren’t confident that everything will go well. That these are the leaders of AI companies racing against each other to make the technology more and more advanced does little to inspire confidence.

(Some of the techno-pessimistic people interviewed for the documentary have expressed their strong displeasure with the final result.)

“Why can’t we just stop?” Roher asks these tech CEOs. He’s told that a moratorium is a pipe dream: Many groups around the world are building advanced AI, all with different motivations. Legislation lags far behind the rate of technological progress. Even if we could pass laws in the US and EU that would stop or slow things down, says Anthropic CEO Dario Amodei, we’d have to convince the Chinese government to follow suit.

If we don’t create it, the thinking goes, our enemies will. It’s best to get ahead of them.

This is, of course, the logic of nuclear deterrence: If we don’t mitigate the risk of ending the world through mutually assured destruction, there’s nothing stopping someone else from pressing the button first.

An apocalypse in every generation

The atomic comparison is apt, if only because Roher sees the stakes in similarly stark terms. “Will my son live in a utopia, or will we go extinct in 10 years?” he wonders aloud. It’s a question that’s central to the film. But he never really sits with the more likely scenario that AI will neither lead to human extinction nor end all disease and drudgery. Every generation faces the specter of its own annihilation — and yet the ends of days keep accumulating, no matter how close the doomsday clock gets to apocalypse.

The point, then, isn’t that AI won’t be bad for us, but that by framing the question in strictly utopian or dystopian terms, we miss the messy reality that lies between hell on earth and heaven in the stars. Although The AI Doc tries to chart an “apocaloptimist” course between two extremes, it doesn’t grasp the real stakes. AI doesn’t really create new risks as such — it’s a force multiplier for existing ones like the threat of nuclear warfare and the development and use of biological weapons. The chief existential risks of AI are human-made and human-driven. And that means, as Caroline says in the film’s ending narration, “We get to decide how this goes.” She’s right, but her husband never seems to understand how she’s right.

Like too many Big Issue Documentaries, Roher’s film is heavy on problems and light on solutions. It does offer some, calling for international cooperation, transparency, legal liabilities for companies if something goes wrong, testing before release, and adaptive rules to match the speed of progress. But just as this is a strictly introductory course in AI — one that will probably irritate those who’ve already moved on to AI 102 — these recommendations are only a starting point. For Roher, they offer reason to be hopeful. For the rest of us, they’re just the beginning of an opportunity to meaningfully steer the course of our future.

Human bodies aren’t ready to travel to Mars. Space medicine can help.

2026-04-07T17:10:14-04:00

Sarah Jane Pell dives with a spacesuit, in 2016 in Marseille, France, simulating lunar conditions. | Alexis Rosenfeld/Getty Images

Editor’s note, April 7, 2026, 5:10 pm ET: The Artemis II mission is conducting experiments that may radically advance our understanding of space medicine. The findings of A Virtual Astronaut Tissue Analog Response (AVATAR) experiment could help us create personalized medical kits for astronauts, and the Artemis Research for Crew Health & Readiness (ARCHeR) study will monitor the astronauts’ health as they go further into space than any human beings have gone before. As we await the findings of those experiments, Vox is republishing this article, which originally launched September 24, 2025.

Vox Members got to read this story first. Support independent journalism and get exclusive access to stories like this by becoming a Vox Member today.

One day, Mars might become a home to humans. But first, there’s the cinematic, sci-fi challenge of making the Red Planet suitable for life. There’s a problem, though: The typical person can’t get to space safely. That throws a wrench into the whole “let’s move to Mars” plan in the face of extreme climate change and other existential risks on Earth.

Today, the path to becoming an astronaut is “littered with the hopes and dreams of medically disqualified candidates,” said Shawna Pandya, a research astronaut with the International Institute for Astronautical Sciences (IIAS) and the director of its Space Medicine Group. “Once upon a time, kids being diagnosed with Type 1 diabetes in the doctor’s office would be told, ‘Well, you could still be anything, except an astronaut.’”

Have space aspirations?

Here are some of the common reasons why you might be medically disqualified from becoming an astronaut:

Tobacco use
Autoimmune disorders
Temporomandibular joint (TMJ) disorders
Sleep apnea
Asthma
Hypertension
Migraines
Anxiety and depression

Astronauts inherently aren’t representative of the broader population — they’re selected for being in very good health. The stress of existing in essentially weightless microgravity conditions, like those on the International Space Station (ISS), can be incredibly tough on the human body. Astronauts face heightened risks of early-onset osteoporosis, insulin resistance, and significant muscle mass loss. Naturally, government space agencies want people whose bodies are more resilient to such pressures, and who can perform necessary duties without a ton of medical intervention.

According to Haig Aintablian, director of the UCLA Space Medicine Program, “just as pregnancy causes the body to undergo complex and unique changes, spaceflight also produces distinct and significant physiological changes.” It also requires its own medical specialty to manage (aptly called space medicine).

There’s a lot scientists don’t know, from the physical to the psychological. That’s a problem — for the future of science, space travel, and maybe even human existence at large.

NASA wants to go to Mars for research, and aims to send humans there as early as the 2030s. As the most similar planet to Earth in our solar system, Mars may have once harbored life, or may even currently. And in the future, we may even need it to support us.

Decades ago, seriously engaging with the idea of moving to Mars was extremely fringe for a multitude of reasons, ranging from a lack of technical feasibility to the desire to put scientific resources toward solving problems on Earth. Elon Musk — founder of the spaceflight company SpaceX — became a famed advocate for colonizing Mars in the early 2000s. He still is. Musk, who is currently worth around $410 billion, claims that he is only accumulating assets for the purpose of Martian space settlement. Last year, he said that he wants 1 million human settlers on the Red Planet in a self-sustaining city by 2050.

Now Musk isn’t alone. NASA experts, biologists, academics, futurists, disaster resilience researchers, and physicians are seriously considering the possibility of making humanity an interplanetary species.

“The biggest problem for humanity to solve is the guaranteed survival of our species — which the logical answer is to become multiplanetary,” Aintablian said. “I don’t think there’s a better solution than Mars.”

While we know some of the health effects of being on the ISS, we can’t really replicate the effects of Martian radiation exposure. Kelly Weinersmith — a biologist and co-author of A City on Mars: Can We Settle Space, Should We Settle Space, and Have We Really Thought This Through? — thinks that settling Mars on Musk’s timescale will be catastrophic. She argues that we shouldn’t rush to set up shop before understanding — and mitigating — the risks, even if this takes centuries rather than decades.

But many advocates for settling Mars are much more impatient. The only way to get there safely would be to unlock significant advances in space medicine, a nascent field that has just barely scratched the surface in its approximately 75-year history.

“Nothing that humanity has done that has been worthwhile has been easy,” Aintablian told me. “So much in our development as a civilization has been difficult, and the reason why we’re able to live such comfortable lives now is because of the extremely difficult challenges that humans have had to solve in the past.”

What we know — and don’t — about human health on Mars

Since extremely few people end up in space right now, the researchers trying to understand how to improve human health there have a limited sample size to work with. Yuri Gagarin became the first human in space in 1961, and more than 600 astronauts have followed him. Only about a sixth of them are women.

NASA researchers have identified some key ways that time in space can impact human health — radiation exposures, isolation, distance from Earth, altered gravity, and environmental consequences like an altered immune system. But we’re still lacking many specific examples of how these different dynamics play out in real life.

Former astronaut Scott Kelly, right, who commanded a one-year mission aboard the International Space Station, along with his twin brother, former astronaut Mark Kelly.

| NASA/AFP via Getty Images" data-portal-copyright="NASA/AFP via Getty Images" />

One of the best studies we have is NASA’s famous 2019 twins study. Twin studies allow researchers to separate the effects of genetic predispositions from environmental influences on health outcomes. NASA compared the health of identical twin brothers Scott and Mark Kelly over the course of a year. Scott went into orbit on the ISS while Mark remained on Earth. Both underwent the same battery of physiological tests, and the results indicated some surprising new differences between the two men.

Scott’s telomeres — the bits of DNA at the end of our chromosomes — lengthened while he was in space and (mostly) reverted to normal once he returned to Earth, possibly indicating radiation-induced DNA damage and potential increased cancer risk. Scott also lost body mass, developed signs of cardiovascular damage that were not present in Mark, and experienced some short-term cognitive changes after returning to Earth.

While survivable with the right training, equipment, and precautions, the twin study demonstrated how space’s unique environment can have significant consequences for gene expression and overall health while in orbit.

If the best of the best struggle, what about the rest of us? We’re getting some insights here now, too.

Since space tourism has literally taken off, astronauts aren’t the only ones going to space now: Wealthy non-astronauts, like Jeff Bezos, Gayle King, and Katy Perry, have recently taken short, recreational jaunts into outer space through Bezos’s space tech company, Blue Origin.

Aintablian is very excited about the prospect of civilian access to space increasing, which will inherently mean people with medical issues are also flying. This represents a huge opportunity for scientists to study the medical management of a much wider range of conditions.

That said, 10 or 15 minutes in space is hardly comparable to the conditions on the ISS. And Mars poses even worse consequences in terms of hostile environments and time spent away from Earth. Mars has toxic dust, lacks plant life and a breathable atmosphere, and only has about 40 percent of Earth’s gravity. Earth’s global magnetic field protects our planet from harmful radiation, and the Martian counterparts are localized, not planet-wide.

The longest time someone has been in space consecutively is 438 days aboard a space station. But crewed missions to Mars would probably take at least nine months just to get there, let alone stay or travel back (which could take up to three years). Mars is usually around 140 million miles from Earth based on its orbital path around the sun, with up to a 20-minute communication delay one way. If they experienced a medical emergency, astronauts likely wouldn’t be able to access telemedicine instructions in time, and they couldn’t turn back around for treatment.

A crewed mission to Mars would have to take all of their supplies with them before they left our planet. And when the first people heading to Mars set foot on the planet, they won’t have access to the intense support astronauts receive when landing back on Earth.

Getting to Mars is only part of the challenge. We’ve been to space, but so far, humans have only ever sent robots to the Red Planet. We are making educated guesses at what Mars is like for living things. Earth analogues aren’t able to truly replicate the closed, hostile conditions of the space environment, which can wreak havoc on astronauts’ mental health. Desert research stations have an atmosphere, while the moon barely has one — and setting up that modest base was a huge mission in its own right. Weinersmith told me that scientists at polar research stations are isolated in remote, inhospitable environments, but they can “still open the door, take a deep breath, and not die.”

Medicine’s new frontier

We’re still pretty far from being able to breathe in Mars’ atmosphere — but it would be nice to get there one day and simply not die.

Enter: space medicine.

Programs dedicated to figuring out how to get humans safely into space for long periods of time are popping up, and non-physician health care providers are getting in on the action too. UCLA is planning to launch a space nursing program and possibly space paramedic training. SpaceMed is a European master’s program focused on human health in spaceflight and other extreme conditions.

Today, astronauts receive most of their care from Earth-based aerospace medicine physicians called flight surgeons through telemedicine. Aintablian envisions a future where health care providers directly accompany astronauts on their expedition-class missions, like to the moon or Mars. Artificial intelligence can act as a resource for the on-board flight surgeon, he predicted, and aid in the development of other technologies that will bring us closer to Mars.

Such technology is already in the works. Google recently collaborated with NASA to develop an AI system that could guide astronauts in diagnosing and treating medical conditions that arise in-flight when they lack access to telemedicine.

But the devil is in the details, Pandya told me. AI can help with just-in-time training for medical emergencies and diagnostics, but the data requirements would be massive. Since extremely few people end up in space — and the ones who do are overwhelmingly male — models might be trained on an unrepresentative dataset that could lead to inaccurate predictions of physiological changes in space. These kinks need to be worked out first.

Right now, there’s a gendered gap in the research — so much so that Weinersmith told me there’s never a line to the women’s restroom at space settlement conferences. Human reproduction and development in space, as a result, is wildly understudied.

As far as we know, no human being has ever been to space while pregnant, and we don’t know of any humans who have been conceived in space. We’re going to learn a lot about reproduction on Earth from the first human space pregnancy and space birth, a prerequisite for a self-sustaining settlement on Mars. (Plus, space tourism companies are talking about hotels in space, and we know what people do in hotels.) Ideally, you want to have an idea of what will happen to someone giving birth in space before they actually go through it.

“What we’re arguing is that we should do the research to understand those risks before we go out there because if there are massive risks, there usually are technological solutions for some of these,” Weinersmith said.

NASA will begin its second Crew Health and Performance Exploration Analog this October, a year-long “mission” to Mars in a 3D-printed habitat at Johnson Space Center in Houston, where it will collect behavioral health data on the effects of isolation and confinement. Scientists are conducting bed rest studies, which simulate the physiological effects of altered gravity and weightlessness. And as funding cuts transform the future of scientific research on Earth and beyond, space medicine researchers are among those advocating for continued investment in space and biomedical science.

Maedeh Mozneb, a biomedical engineer and project scientist in the Sharma Lab at Cedars-Sinai Medical Center, told me that the ultimate goal is to send “avatars” of astronauts to space by taking their stem cells and creating 3D tissue cultures called organoids that represent different parts of their body — yes, miniature hearts, kidneys, and even brains made from Earth-dwelling humans. From there, scientists can determine personalized countermeasures such as workout plans or supplements tailored to each astronaut’s needs, before they actually end up in space.

The hope, for those space medicine physicians like Pandya, is that in a spacefaring future, all medical disciplines — from neurology to radiology — will be represented in space medicine.

Red planet, blue planet

Space medicine research and practice isn’t cheap. “I often get asked,” said Pandya, “‘Why are you spending money on space health when we have all of these problems on Earth?’” But that’s the wrong way to think about it, she said.

Research conducted in space has already improved health on this planet. Advances in digital imaging for moon photography during the 1972 Apollo 17 mission later played a crucial role in CT scans and MRIs. Remote health monitoring tools designed for astronauts in space are now widely used in hospitals.

One of the next big things in space medicine “is probably going to be the development of radiation protection mechanisms,” Aintablian told me.

Space medicine research will also allow more people to go to space. In 2023, Pandya’s team demonstrated the safety and functionality of a continuous glucose monitor in the spaceflight environment. This could eventually allow diabetics to check their blood sugar in space. It has implications for current astronauts, who can develop insulin resistance and pre-diabetes symptoms in longer-duration spaceflights. The child diagnosed with Type 1 diabetes who wants to be an astronaut may actually have the chance to live out their dream now, and studying how the body metabolizes glucose in space helps us better understand health on Earth.

Then there are the diseases that take decades to unfold. Muscle loss in space can help scientists better understand how to treat conditions like Duchenne’s muscular dystrophy. On Earth, neurodegenerative diseases like Alzheimer’s often aren’t apparent until a person is in their late 60s.

In microgravity, said Shelby Giza, the director of business development at Space Tango, a company that facilitates automated research and development in microgravity conditions, “you can see that kind of disease output in a matter of weeks.” Research on these conditions can be conducted much faster — and hopefully accelerate the pace of medical breakthroughs.

The same can be said for cancer. Not all radiation exposures are made equal, and susceptibility to the harmful effects of radiation varies between individuals. Since the ISS is within the protection of Earth’s magnetosphere, it’s not the best comparison to the elevated radiation levels astronauts would face on Mars.

According to former NASA astronaut and biologist Kate Rubins, most astronauts are healthy people in their 30s and 40s, an age when cancer typically doesn’t develop. Scientists must track astronauts for decades after their last spaceflight to see if cancer or other adverse health conditions occur. NASA’s Lifetime Surveillance of Astronaut Health program, which is voluntary for former astronauts and not specific to cancer alone, monitors the health status of people like Kelly and Rubins for the rest of their lives.

Exposure to space radiation is linked to developing cancer and degenerative diseases. To mitigate the risk of developing fatal cancers, NASA currently limits astronauts’ spaceflight radiation exposure to 600 millisieverts (mSv) — roughly the equivalent of 60 CT scans of the torso and pelvis — over the course of their entire career. A 2023 NASA white paper estimates that a healthy astronaut will have a 33 percent increased risk of dying from cancer in their lifetime after a 1,000-day mission to Mars.

One of the next big things in space medicine “is probably going to be the development of radiation protection mechanisms,” Aintablian told me. “I do believe that with the amount of emphasis being placed on radiation protection, we’re going to figure out ways to actually protect against significant amounts of radiation for the general public for multiple uses.”

While it’s still relatively early days for the space pharma industry, life science companies are taking note, seeing microgravity as a platform for better drug discovery.

Like fiber optic cables used for telecommunications, some pharmaceuticals are better synthesized in microgravity conditions. Scientists can produce more uniform protein crystals in microgravity, which can improve drug injectability and reduce the need for refrigeration.

Raphael Roettgen, an entrepreneur and the co-founder of space biotech startup Prometheus Life Technologies, told me that organoids — those 3D cell models replicating human organs — grow more cleanly in space without Earth’s gravity weighing them down. Derived from non-embryonic stem cells, these miniature organ models have tremendous potential for personalized medicine.

Roettgen hopes that human space organoids could reduce the need for animal testing in the near term. Eventually, he hopes that new organs could be regenerated for patients needing transplants. Since the new tissue would be derived from the patient’s own stem cells, there would not be a risk of immune rejection, saving transplant patients astronomical costs and immense suffering. He estimates that liver regeneration and transplants from these organoids could become a reality in patients within the next 20 years.

Microgravity is an “expensive tool,” but an important one nonetheless, said Mozneb, who studies the effects of low earth orbit on stem cell differentiation. She hopes increasing commercialization and new technologies will significantly decrease the cost of launching experiments into orbit over the next 10 years.

What we already know about space medicine is a drop in the ocean of what we will discover as more people — astronauts and otherwise — venture into space.

“It’s like if you were studying genetics back in the ’90s,” Mozneb said. “Everything is a discovery.”

The world’s deadliest infectious disease is on the rise in the US

2026-03-24T13:20:02-04:00

Various types of bacteria, 1889. | Oxford Science Archive/Print Collector/Getty Images

Something unusual happened at Archbishop Riordan High School last fall.

In September, a student in the Bay Area school went to see a health care provider for a cough that wouldn’t go away. But it wasn’t until two months later that the student got diagnosed: tuberculosis. The San Francisco Department of Public Health (SFDPH) launched an investigation, which revealed a surprisingly high rate of latent tuberculosis — meaning that people were infected by the Mycobacterium tuberculosis bacteria, but their infections had not yet progressed to active and contagious disease — at the school.

As of February 24, the most recent data available, four people in the school community had confirmed active tuberculosis, and an additional three active cases were suspected by the public health department.

A private school in San Francisco isn’t exactly where you would expect a tuberculosis outbreak to occur. Tuberculosis is largely a disease of poverty and marginalization, and today the developing world bears the greatest burden. The vast majority of all new cases (about 87 percent) occur in just 30 low- and middle-income countries.

But it used to be far more prevalent globally. Rewind the clock: On March 24, 1882, a German physician named Robert Koch announced that he had identified the cause of the illness that killed one out of every seven people in the US and Europe. Now fast-forward: Today is World Tuberculosis Day, marking the 144th anniversary of Koch’s discovery. And the disease is making a comeback in wealthy countries.

Call it consumption, “the robber of youth,” the white plague — but we certainly can’t call it gone. And although it was briefly outpaced by Covid-19, in 2023 tuberculosis regained its title as the world’s leading cause of death by infectious disease. Every year, it infects about 10 million people and kills 1.5 million — despite being both preventable and curable. Counting both latent and active cases, a fourth of the entire human population may be infected with the bacteria worldwide.

“The global is local and the local is global, so if we’re not able to address the global burden of tuberculosis, we’ll continue to see it everywhere,” Priya Shete, an associate professor of medicine and tuberculosis researcher at University of California San Francisco, told me. “We’ll start to see tuberculosis arise in the least expected places.”

The United States has the infrastructure for tuberculosis testing and treatment, and it isn’t currently endemic here. Like much of the world, it used to be though — it may have killed as many as a quarter of all Americans during parts of the 18th and 19th centuries. But improvements in nutrition, living conditions, sanitation, and, especially, the advent of antibiotics in the mid-1900s changed things dramatically. Still, “not endemic” is a far cry from “eradicated.”

After 30 years of being on the decline, tuberculosis rates are rising in the US. In February alone, it popped up in American high schools beyond the Bay Area, with confirmed cases in Long Island, New York, and Seattle. One of the largest American outbreaks since the Centers for Disease Control and Prevention started reporting tuberculosis data in the 1950s happened just two years ago in Kansas, leading to 68 active cases, 91 latent infections, and two deaths.

Broad disinvestment in public health infrastructure, medication supply shortages, delays in diagnosis due to the Covid pandemic, and the challenges of detecting and treating latent cases are all part of why tuberculosis is still a disease worth worrying about in the United States.

The theme of this year’s World Tuberculosis Day is “Yes! We Can End TB!” That’s very ambitious, given that it’s still an ongoing challenge even in the world’s richest nation. Its persistence requires us to stay ahead in the evolutionary arms race with the pathogen, one that has possibly been on Earth for 3 million years. But there is hope — advocates are pushing for sustained investment in tuberculosis research and fighting back against funding cuts, and scientists are working to develop new treatments for this very old disease.

The US v. tuberculosis

Tuberculosis spreads through the air, and a cough isn’t required to infect someone else — regular breathing can do the trick. “There is no country in the world that has eliminated TB or that is TB free,” Lucica Ditiu, a physician and the executive director of the Stop TB Partnership, told me. “As long as we breathe, we are all at risk.”

Approximately 13 million people in the US have latent tuberculosis, which progresses to active disease in about 5 to 15 percent of cases when left untreated. Tuberculosis preventative treatment (TPT) for latent cases can clear up the infection in as little as three months with appropriate antibiotics, but a lack of access to these medications or delays in diagnosis often prevent patients from starting TPT.

Left untreated, someone with active tuberculosis can infect 15 people a year.

Latent infections can become active when the immune system is no longer able to keep the bacteria in check. Babies, small children, the elderly, and immunosuppressed people are at the greatest risk of converting from latent to active disease. Our health is not constant over the course of our lives: “You can have latent TB your entire life, and in your last chapter…when your immune system is suppressed, it can become active and you can start spreading it,” RC Sadoff, a public health researcher at Johns Hopkins University and team member of Zero TB in Kids, which provides tuberculosis screening and treatment for children in South Asia in settings where people live in close proximity like boarding schools and monasteries. “The powder keg is already everywhere.”

Left untreated, someone with active tuberculosis can infect 15 people a year. In the US, more than 80 percent of cases result from latent cases becoming active. To complicate matters, the distinction between latent (and non-infectious) and active (and contagious) tuberculosis may not be a binary but a continuum. It’s not always clear at what point someone becomes infectious, and people can be contagious without having the traditional tuberculosis symptoms. One study estimates that subclinical cases — active infection but symptom-free — might account for 68 percent of tuberculosis transmission around the globe.

“I think it is plausible that at least some of the latent TB burden we are seeing in [the Archbishop Riordan High School community] could have been transmitted by people without symptoms but still infected with tuberculosis,” Sadoff said. “I think if we understood more about subclinical transmission, maybe we could intervene sooner or more effectively [in outbreaks like these].”

Compared to something exponentially more infectious like measles, tuberculosis is much more insidious. The main symptom people generally have is a chronic cough.

“It gets diagnosed as everything but tuberculosis, and it goes on for a long time. So I think the danger of tuberculosis is that it can look and seem like a lot of other things that aren’t so bad,” Shete told me. When people have night sweats, another common symptom of tuberculosis, they’re more likely to think it’s hot flashes than tuberculosis, she said. “It takes a bit more vigilance to think about TB.”

Delays in TB diagnosis were exacerbated by Covid, and that impact still has not been reversed. Today, new tuberculosis patients are often sicker by the time they’re diagnosed than they were in pre-pandemic caseloads. Delays in diagnosis and treatment mean more time for the disease to spread unabated, a greater chance that the bacteria will develop antibiotic resistance, and more preventable suffering. Successful tuberculosis treatment also requires sticking to a schedule of several antibiotics for months, so that the bacterium does not develop resistance to the medication.

Delays in TB diagnosis were exacerbated by Covid, and that impact still has not been reversed.

Another major barrier to combatting the disease in the US is cost. According to the California Department of Public Health, TPT for latent tuberculosis can cost one patient around $857 for a three- to four-month course of treatment. Diagnosing and treating a patient with active tuberculosis costs more than 50 times that. If TB and multidrug-resistant TB in the US were to reach current global average rates, the cost of treating TB cases in the US would increase to over $11 billion annually.

Funding cuts to programs like USAID could lead to 2.2 million additional tuberculosis deaths in 26 countries with a high burden of the disease between 2025 and 2030. Drug-resistant cases could increase by 30 percent, requiring medications that are hard to come by and may not be FDA-approved for use in the US. We’ll feel that at home; it can cost $154,000 to treat and cure one person with drug resistant tuberculosis in America.

In the US, the disease typically spreads in places without sufficient health care services to diagnose people with active TB and in places that are overcrowded, have poor ventilation, and higher rates of malnutrition like prisons and homeless shelters. As in other parts of the world, the most marginalized in the US are at the greatest risk — and treating them is more difficult.

“It takes a weak health system in a high-income country to produce a tuberculosis outbreak,” Sadoff told me. “Whenever there’s a TB outbreak in America, something is wrong.”

Preventing the spread of TB in the US — and beyond

At Archbishop Riordan High School, 207 members of the school community have tested positive for latent tuberculosis. In order to return to in-person instruction, students had to share their test results, what treatment they were on, and how long they had taken medication, but they didn’t have to test negative for the bacteria. That’s according to a grandparent of a student infected with latent tuberculosis, who spoke to me anonymously out of fear of retaliation from the school.

About 80 percent of students with latent tuberculosis were on TPT as of February 24. The school cannot legally require students with latent tuberculosis to take the medication — even though it might prevent the infection from progressing to active disease. But, the grandparent said, the school chose not to offer continued online learning options for students who aren’t receiving treatment: “To me, that’s not the way to control it or keep it from spreading.”

Archbishop Riordan High School did not respond to multiple requests for comment on this policy or to several other specific questions about the outbreak. Everyone either confirmed or suspected to have active tuberculosis is on treatment and no longer considered contagious, according to the latest in a series of “TB town hall” webinars hosted for the school community.

“Individuals with latent TB infection (LTBI) are allowed on campus because LTBI is not contagious,” the San Francisco Department of Public Health told me over email. It “strongly recommends that all individuals with LTBI begin and complete treatment, and we continue to educate the community about its importance in preventing future disease.” Once someone has tested positive for tuberculosis, the test will remain positive indefinitely even after successful treatment, which poses a challenge for diagnosis and future surveillance: People who are cured can be reinfected.

The Archbishop Riordan High School story shows how containment anywhere is a complicated and contentious project. No matter where a tuberculosis outbreak occurs, reassuring a community that it is safe is challenging. That’s a major reason why prevention is the ideal.

Health officials from the public health department stress that the risk to the broader community is low, and we don’t currently see much community transmission in the US. But the further gutting of US public health infrastructure and research funding jeopardizes our ability to keep the incidence of this disease so low.

If you’d like to do something about this global threat, you can donate to Stop TB Partnership, TB Alliance, The Global Fund, Partners in Health, and Spiro, which provides screening and TPT for children in Pakistan, a country with exceptionally high rates of the disease. You can become a TBFighter, part of author John Green’s collective to address the structural causes of tuberculosis. You can also become a member of Stop TB USA, which makes it easy to contact your representatives to advocate for continued funding for public health. They also accept donations.

Developing new medications, better diagnostics, vaccines that protect adults from transmission, and advocating for sustained funding into tuberculosis research, prevention, and treatment are the most powerful tools to someday end TB in the US and beyond. Even if it doesn’t happen before next year’s World Tuberculosis Day.

NASA’s first medical evacuation is here. It won’t be the last.

2026-02-26T16:45:59-05:00

Crew-11 mission astronauts pause outside the Neil A. Armstrong Operations and Checkout Building en route to launch complex LC-39A at the Kennedy Space Center in Cape Canaveral, Florida on August 1, 2025. From left: Roscosmos cosmonaut Oleg Platonov, NASA astronaut Mike Fincke, NASA astronaut and mission commander Zena Cardman, and JAXA astronaut Kimiya Yui. | Gregg Newton/AFP via Getty Images

The first medical evacuation in the history of the International Space Station (ISS) happened last month.

In January, Crew-11 returned to Earth ahead of schedule because of an unspecified medical issue. NASA initially didn’t reveal which astronaut was dealing with the problem, citing privacy concerns, but indicated that the person’s condition is stable.

On Wednesday, NASA astronaut Mike Fincke revealed that it was his condition that prompted the crew’s early return, though he did not elaborate on the nature of his medical issue. In addition to Fincke, the group included NASA astronaut Zena Cardman, Russian cosmonaut Oleg Platonov, and Japanese astronaut Kimiya Yui. The reason why the whole crew had to return home (and in the SpaceX capsule they came from) was because there were no spare crew-ready capsules at the moment, and NASA wanted to avoid leaving astronauts in orbit without a way back. Crew-11, which left for the ISS in August, was nearing the end of its six-month mission anyway, making the call a bit simpler.

The ISS, which originally launched in 1998, has been continuously occupied by rotating crews of astronauts since late 2000, and it serves as an important international laboratory for developing new technologies and medicines, as well as studying life in the space environment. However, Crew-11’s departure doesn’t mean the ISS is empty; it was staffed by a skeleton crew of three until Crew-12 arrived in mid-February.

NASA’s chief health and medical officer James Polk said that the medical issue was not an injury sustained while performing work on the ISS but, rather, a health concern arising in the microgravity environment.

On January 7, “I experienced a medical event that required immediate attention from my incredible crewmates,” Fincke wrote in his February statement. He was quickly stabilized, and NASA decided that the crew should return early to “take advantage of advanced medical imaging not available on the space station.”

“Everyone on board is stable, safe, and well-cared for,” Fincke wrote in a LinkedIn post from the ISS before the evacuation. “This was a deliberate decision to allow the right medical evaluations to happen on the ground, where the full range of diagnostic capability exists. It’s the right call, even if it’s a bit bittersweet.”

This is also the first time in NASA’s history that a mission has ended early because of a medical issue. It’s not the first time ever; the Soviets performed two medical evacuations for cosmonauts in the 1980s. According to Polk, statistical models suggest that there should be a medical evacuation from the ISS about every three years, but it’s been smooth sailing for the past quarter-century.

“It’s almost amazing that we’ve maintained the ISS for [almost] 26 years constantly crewed without something like this happening before,” Jordan Bimm, a historian of US space exploration at the University of Chicago, told Scientific American.

Medical evacuations have been kept to a minimum, likely in part because all astronauts receive medical training and have access to telemedicine services on the ground. Several medical issues that arose in-orbit have been dealt with appropriately; no evacuation was required. One astronaut who developed a blood clot on the ISS, for example, was successfully treated with blood thinners. It’s a great reminder as to why innovations in the field of space medicine are so important. As more and more people leave for space, we’ll have to figure out how to treat them without requiring a return trip back to Earth.

Why space medicine matters

Space is obviously tough on the body.

In the microgravity environment — like one you would find on the ISS — astronauts face greater risks of early-onset osteoporosis, insulin resistance, and significant muscle loss. And without normal gravity exerting its pull on the body, blood and other bodily fluids shift “up” toward the head, reducing the volume of blood circulating through the heart and blood vessels. The heart slacks off, becoming rounder. Cardiac muscles that normally work to constrict blood vessels atrophy on the ISS, and these changes become more pronounced over time.

Pressure changes and fluid shifts in microgravity can lead to spaceflight-associated neuro-ocular syndrome, which causes swelling and changes to the shape of the eye and brain. It can make astronauts’ famously 20/20 vision blurry. The longer astronauts are in space, the more likely they are to experience symptoms. While these changes (usually) revert when astronauts return to Earth, they could be more permanent in some people. And in the long term, exposure to space radiation is linked to developing cancer and degenerative diseases.

All of this sounds pretty grim if you want to spend the long stretches in space that would enable us to travel to Mars for scientific research or establish permanent off-world human settlements. But it’s not all bad news. Space medicine works to safeguard human health in the space environment, and it has tremendous benefits for the rest of us here at home.

Last year, I reported on space medicine and its promise to transform health on Earth. Haig Aintablian, the director of the UCLA Space Medicine Program, told me that one of the next big things in the field “is probably going to be the development of radiation protection mechanisms. I do believe that with the amount of emphasis being placed on radiation protection, we’re going to figure out ways to actually protect against significant amounts of radiation for the general public for multiple uses.”

AI could also act as a resource for the on-board flight surgeon, he predicted. It’s not so far off; Google collaborated with NASA to develop an AI system that could guide astronauts through diagnosing and treating medical ailments that arise in-flight, which could improve off-planet (and thus potentially on-planet) diagnostic capabilities.

In terms of this evacuation, “this seems abnormal now, but it is a preview of what will be the new normal if humans go to space in greater numbers,” Bimm told Scientific American. “People will get sick, and sometimes contingencies will have to be exercised.”

NASA livestreamed the departure of the capsule starting at 4:45 pm ET on January 14. Crew-11 landed safely in the Pacific Ocean off the coast of California in the early morning the following day. You can watch NASA’s coverage of the event on the agency’s YouTube channel.

Update, February 26,2026, 4:45 pm: This story, originally published January 14, 2026, has been updated after new information became public about details of the issue that led to the evacuation.

Is it love? Or is it an AI romance scam?

2026-02-13T16:10:15-05:00

Happy Valentine’s Day. Don’t let romance scams — which ramp up around the holiday and are at an all-time high — break your heart.

These scams cost Americans $3 billion last year alone. That’s almost certainly an undercount, given victims’ particular reluctance to report that they’ve fallen for such ruses.

Many romance scams fall under the umbrella of so-called “pig-butchering” scams, in which fraudsters build relationships with and gain the trust of victims over long periods of time. The moniker is a crude reference to fattening up a pig before the slaughter — and they go for the whole hog, repeatedly attempting to extract money from the target. Between 2020 and 2024, these scams defrauded more than $75 billion from people around the world.

Now, AI is making these scams increasingly accessible, affordable, and profitable for scammers. In the past, romance scammers had to have a strong grasp of the English language if they wanted to effectively scam Americans. According to Fred Heiding, a postdoctoral researcher at the Harvard Kennedy School who studies AI and cybersecurity, AI-enabled translation has completely removed that roadblock — and scammers now have millions more potential victims at their disposal.

AI is fundamentally changing the scale, serving as a force multiplier for scammers. A single person who used to manage a few scams at a time can use these toolkits to run 20 or more simultaneously, Chris Nyhuis, the founder of cybersecurity firm Vigilant, told me over email. AI-assisted scams are significantly more profitable than traditional ones, and they’re increasingly cheap and easy to run.

On the dark web, fraudsters can purchase romance scam toolkits complete with customer support, user reviews, and tiered pricing packages. These toolkits come with pre-built fake personas with AI-generated photosets, conversation scripts for each stage of the scam, and deepfake video tools, Nyhuis told me. “The skill barrier to entry is essentially gone.”

I wondered if romance scammers might automate themselves out of a job, but the Kennedy School’s Heiding told me that “oftentimes it’s just augmentation, rather than complete automation.” Many of the scammers are also victims themselves, with at least 220,000 people trapped in scam centers in Southeast Asia and forced to defraud targets, facing terrible abuse if they refuse. Leveraging AI means “the crime syndicates [who run these centers] will probably just have better profit margins,” Heiding said.

For now, there’s a human being behind the scenes of the scams, even if they’re just pressing start on an AI agent. But apart from that, it can be fully automated. At the moment, Heiding told me, AI isn’t much better than human romance scammers, but the technology evolves rapidly. In 2016, Google DeepMind’s AlphaGo beat the world’s best human go player in a landslide. Human forecasters think that AI is set to far outpace their ability to predict the future very soon.

“I wouldn’t be surprised [if] within a few years or a decade, we have AI scammers that are just thinking in completely different patterns than humans,” Heiding said. “And unfortunately, they probably will be really, really good at persuading us.”

What’s love got to do with it?

Romance scams are unique: They target a core human need for love and connection. You may have heard that we’re in a loneliness epidemic, officially declared by the US Surgeon General in 2023, with health risks on par with smoking up to 15 cigarettes a day. Social isolation is linked to higher rates of heart disease, dementia, depression, and even premature death – and reportedly, 1 in 6 people worldwide are lonely. And lonely people make for prime targets.

Fraudsters send out initial AI-generated messages to prospective victims. Over time, they use lovebombing techniques to convince them that they are in a romantic relationship. Once trust is established, they make requests for money through methods that are difficult to recover like gift cards, wire transfers, or cryptocurrency. They will often make up crises that require urgent transfers. They might ghost the victim after reaching their goals, or continue the scam to squeeze more out of them.

AI romance scams use deepfake video calls, “cheap fake” social media profiles, and voice cloning technology like other AI-enabled scams to draw people in. But according to Nyhuis, they’re “uniquely dangerous because of what they exploit. Phishing uses urgency; tech support scams use fear. Romance scams use love, which can make people think irrationally or overlook their gut feeling that something is wrong.”

Older adults often experience social isolation and are frequently targeted by romance scammers. Retirement and bereavement can create circumstances that scammers deliberately manipulate, making victims feel seen and cared for, even as they steal their life savings and the homes where they plan to spend their retirement years. But anyone can be deceived by these scams. Despite being digital natives, Gen Z is three times more vulnerable to online scams than older generations since they spend so much time online, although they tend to have — and therefore lose — less money than older victims.

Here’s something else that will break your heart: Scam victims are more likely to be targeted again. Scammers create profiles of their targets, sometimes adding them to “sucker lists” shared across criminal networks. Victims of other crimes are also more likely to be revictimized, and falling prey to a romance scam isn’t a moral failing on the part of the target.

But it is something to be on guard against, since the vast majority of scam victims will not be able to get their money back. About 15 percent of Americans have lost money to online romance scams, and only 1 in 4 were able to recover all the stolen funds.

Romance scams thrive in shame and secrecy. Victims are sometimes blackmailed and told that if they confide in people in their lives, the scammers will expose sensitive information. Sanchari Das, an assistant professor and AI researcher at George Mason University, and Ruba Abu-Salma, a senior lecturer in computer science at King’s College London, received a Google Academic Research Award to study AI-powered romance scams targeting older adults in 13 countries. Their research examines how AI tools can amplify traditional scam tactics and how families and communities can better support the victims.

The researchers are building connections with gerontological societies, and aim to build educational tools to support AI romance scam victims. There’s a fair amount of information already out there about prevention, but very little directing victims what to do next.

Love in the time of AI

Like so many people, I met my partner online. I’m grateful that we started dating in the late 2010s, before the explosion of AI-generated profiles on apps and dating sites.

AI is getting better at tricking people across the board. It has massively improved at rendering hands, a formerly reliable tell for deepfakes, and it learns from its mistakes. “As these technologies improve, traditional signals for spotting manipulation are no longer dependable,” Das said. “At the same time, we are leveraging AI to counter these threats by detecting scam patterns, forecasting emerging tactics, and strengthening protective responses. The goal is to build systems and communities that are as adaptive as the technology itself.”

Society is also getting increasingly desensitized to AI romance. One study found that almost a third of Americans had an intimate or romantic relationship with an AI chatbot. The 2013 movie Her, in which a man falls in love with an AI voiced by Scarlett Johansson, was set in 2025. It wasn’t too far off the mark.

AI chatbots are purposefully designed to keep people engaged. Many use a “freemium” model, in which basic services don’t cost anything, but charge a premium for longer conversations and more personalized interactions. Some “companion bots” are designed to make users form deep connections. Even though people know that the “significant other” is AI, these companion bot apps sell user data for targeted advertising and aren’t transparent about their privacy policies. Is that not also a sort of intimacy scam, a way to extract resources from lonely people for as long as possible?

There are steps you can take to protect your heart, wallet, and peace of mind. It seems obvious, but refusing to send money to someone you haven’t met in person will stop a romance scam in its tracks. You can demand spontaneous video calls, and ask the person on the other end to do something random; deepfakes still struggle with “unscripted” actions.

“Be suspicious of anyone you’ve never met in person — that’s the only safe approach in a digital world increasingly filled with scams,” Konstantin Levinzon, the co-founder of free VPN service provider PlanetVPN, said in a press release. “If someone you meet on a dating site seems suspicious, perform a reverse image search to check if their pictures are stolen from other sources. And if the conversation shifts to money, or if someone asks for personal information, leave the conversation immediately.”

You can also use a VPN to obscure your location, since scammers might track users’ location and try to personalize their scams based on the target’s city or country. If you are scammed, reporting early to the FBI Internet Crime Complaint Center, Federal Trade Commission, and your bank increases the chances that you’ll be able to recover the stolen funds. Several nonprofits offer support for victims of romance scams.

“No matter how alone you feel right now, no matter how embarrassed you are, you will recover from this and one day look back and see how you made it through it,” Nyhuis said. “These scammers are good at removing hope. Don’t let them take that from you.”

The public health win hiding in plain sight: Poison centers

2026-01-27T15:45:14-05:00

Accidentally poisoning yourself is much easier than you might think.

Once, I took two antihistamine medications on the same night and panicked when Google told me that there might be an interaction between them. So, I called the toll-free Poison Help Line — 1-800-222-1222, listed on many household chemical bottles — and they told me that I was fine, saving me from an unnecessary emergency room visit and needless stress.

This story was first featured in the Future Perfect newsletter.

Sign up here to explore the big, complicated problems the world faces and the most efficient ways to solve them. Sent twice a week.

Beyond reassurance, the US Poison Center Network — a constellation of 53 regional centers serving every state and territory — offers lifesaving first aid guidance and useful data surveillance. Staffed by medical professionals trained in toxicology, many centers now also offer text and live chat services in multiple languages in addition to the traditional 24/7 helpline. From your grandma unwittingly eating your marijuana-laced brownie to your child swallowing an unknown pill, poison control is there to walk you through what to do next. Many accidental poison exposures are able to be safely managed at home with expert assistance.

In the past 30 days alone, there have been 201,545 reported poisonings in the US, according to the National Poison Data System.

With so many poisonings happening all of the time, the Poison Center Network is incredibly valuable. In fact, it saves about $3.1 billion every year in health care and productivity costs, according to a new report from RAND, a nonprofit policy think tank. It found that, for every dollar invested, American communities get $16.77 in benefits from lower emergency department use, less time spent in the hospital, better health outcomes, and lower risk of death.

Poison centers are an undersung public health win — a model that has worked, and evolved, over the past 70 years, even as Google and AI become many people’s first go-to for information, even in a crisis.

Hollowing out a vital lifeline

However, recent budget cuts threaten poison centers’ ability to carry out their lifesaving mission, and federal and state funding has not been adjusted for inflation in over a decade.

Poison control centers depend significantly on federal funding sources like Medicaid and the Children’s Health Insurance Program, which have faced significant cuts in recent years. The average operating costs for a poison center ranges from $1.2 million to $7.2 million, excluding subsidized support. All together, federal funding for poison centers saves $450 million in health care costs alone annually, making them a pretty clearly good investment on the math alone — not to mention the lives and suffering saved.

The network also contributes an incredible wealth of health and safety information. Its crown jewel, its data system, is the only near-real time public health surveillance system of its kind in the US, with data uploaded every 4.97 minutes. It’s helpful for that data to be as up-to-date as possible, because one poison exposure is reported to a center every 15 seconds.

And their mission has expanded. Poison centers have taken on a greater role in emergency preparedness and response, and many provide additional functions like operating a rabies and Covid-19 hotline, conducting research, and providing telehealth delivery. These “ancillary functions” can generate revenue for the centers providing them through government or industry contracts, helping them to cover operational costs, but they require the centers to offer additional services on top of their core toxicology work. Even with the current coverage, more than 100,000 people in the US died from preventable poisonings in 2023.

Since 2000, poison centers have averaged more than 3.3 million encounters each year. While total touchpoints have declined since the 2010s, probably because of new online information sources, the average severity of cases has increased. The report found that 30 percent of human exposure cases came from a health care facility or provider contacting poison control, suggesting that poison centers are spending more time and resources on the cases that come to them.

It can be hard to access health care services, and people need accurate and actionable information in a crisis. And with potentially unreliable and unvetted information online, poison centers, staffed by trained professionals, are a lifeline.

If you want to find your local poison center and find out how you can support them, click here.

50,000 clandestine Starlink terminals are keeping Iranians in touch with the rest of the world

2026-01-16T16:56:54-05:00

Security forces are seen during a pro-government rally on January 12, 2026, in Tehran, Iran. | Stringer/Getty Images

It’s difficult to know exactly what is happening in Iran since the government shut down the internet on January 8, plunging a nation of more than 90 million people into digital darkness.

Crackdowns against anti-government protesters have led to at least 2,600 deaths, although some estimates put the death toll at upward of 20,000. According to the US-based Human Rights Activists News Agency, more than 18,000 protesters have been arrested.

The protests began in late December in response to dire economic conditions and took on a broader anti-government character as people demanded the end of Ali Khamenei’s rule. The Iranian rial is now the least valuable currency in the world. The country has an inflation rate of about 40 percent, making necessities unaffordable for most people. Iran is struggling through a long-lasting economic crisis, driven by sanctions, government austerity measures, and last year’s war with Israel. Many parts of the country, including the capital of Tehran, face severe and unrelenting drought, as I reported in November.

The government also cut phone lines on January 8. While the government eased some of these restrictions on Tuesday, allowing some Iranians to make international calls out of the country this week, many reasonably fear government surveillance. People outside the country remain unable to call Iranians. Several people in Tehran called the Associated Press on Tuesday, saying that text messaging services remain down and that internet users could connect to local government-approved websites but not to international ones.

So Elon Musk’s Starlink — which provides high-speed internet access in difficult-to-reach places via satellites that receive radio signals from user terminals on the ground — has become a lifeline for Iranians trying to share what is happening on the ground. SpaceX has made Starlink free for its tens of thousands of Iranian users, but since the Iranian government criminalized the use of satellite internet services like Starlink last year, they face substantial risk in accessing it illegally.

And yet many Iranians are using it anyway.

If satellites are in jeopardy, so is the truth itself.

According to Iranian internet rights group Filter.Watch, the government has attempted to jam signals from Starlink satellites and is actively hunting down people they believe to be using the service.

New updates to the Starlink terminals thwarted some of the government’s efforts to jam the signal. Since Starlink launched in 2022, activists have smuggled terminals into the country, and there are now about 50,000 hidden in the country. Developers have created tools to share Starlink connections beyond a single terminal.

“A big problem with Starlink is that ultimately it represents a single point of failure for communications,” Steve Feldstein, a political scientist and senior fellow at the Carnegie Endowment for International Peace, told me over email. Despite this, Starlink is the best option Iranians have. “No other tool provides as much scalability and affordability to Iranian citizens,” Feldstein said.

At a time when disinformation and intentional obfuscation can downplay the scale of death or hide that atrocities are occurring at all, satellites — and not just Starlink’s — are proving their place in uncovering humanitarian crises. Without them, the world will be left in the dark.

Satellites are a human rights issue

Satellites are effectively the only way to follow humanitarian crises during information blackouts or when no one can get in or out. In November, my colleague Sara Herschander reported on the Sudanese civil war, in which the violence is so severe the bloodshed is visible from space. Only satellite imagery and geolocated social media posts provided evidence of the atrocities due to a communication blackout.

Around 15,000 satellites currently orbit the Earth; the number has rocketed up in recent years as companies launch large satellite networks called megaconstellations to provide broadband internet access. Most of them are in low Earth orbit, up to 1,200 miles above the Earth’s surface. More than two-thirds of active satellites in low Earth orbit belong to the Starlink megaconstellation.

Bear with me for a second, but if you care about what’s happening on Earth, there’s one thing we have to worry about: space traffic.

By 2040, there will be more than 560,000 satellites in orbit. The more satellites we send up, the greater the risk that they will collide into one another or bits of space junk. This could lead to massive service disruptions, or in the worst case, lead to a phenomenon known as Kessler syndrome. That’s when a cascade of new collisions happens in a chain reaction, potentially rendering low Earth orbit unusable — meaning no more satellite launches, an end to our space exploration ambitions, and the severe disruption of technologies like GPS, weather alerts, and satellite internet.

But that’s a worst-case scenario, and SpaceX is aware of it. The company announced on January 1 that it plans to lower 4,400 of their satellites from 342 to 298 miles above the Earth’s surface over the course of the year to reduce collision risks.

In 2023, the United Nations’ International Telecommunications Union estimated that 2.6 billion people — a third of humanity — lack internet connectivity. The UN considers internet access to be a human right. An underappreciated consequence of low Earth orbit becoming increasingly unusable is losing satellite internet access and imagery that allows us to see past rhetoric.

Satellite imagery is how we know what is happening in conflict zones like Ukraine and Sudan. If satellites are in jeopardy, so is the truth itself.

26 things we think will happen in 2026

2026-01-06T15:26:47-05:00

For the seventh year in a row, the Future Perfect staff — plus assorted other experts from around Vox — convened near the end of the year to make forecasts about major events in 2026.

Perhaps in keeping with the year we just experienced, the prognostication had grim overtones. Will the US remain an electoral democracy? Will the country fall into a recession? Will there be war in Taiwan? Will more states ban lab-cultivated meat? Will a Category 5 hurricane make landfall in the US? Will Beyoncé release a rock album? (Which is maybe just grim to me — there are so many better options!)

As always, we try to avoid random guessing. Each prediction comes with a probability attached. That’s meant to give you a sense of our confidence in our forecasts. The idea here is to exemplify epistemic honesty — being as transparent as we can about what we know we know, what we know we don’t, and what we don’t know, we don’t know.

As we have every year, we’ll check back at the end of the year and provide a report card on how we did, whether our accuracy ends up being Nostradamus level, or more like a band of blindfolded monkeys throwing darts at a board. You can check out how we did in 2025 here. We hope you enjoy reading — and don’t forget to update your priors. —Bryan Walsh

The United States

The US falls from the ranks of liberal democracies in the leading V-DEM index, but remains an electoral democracy (60 percent)

Entering 2026, assessing the health of American democracy is a bit of a puzzle.

There is no doubt that, in the first year of Donald Trump’s second presidency, American democracy has weakened significantly. He has smashed through constitutional constraints on his power, targeted his political opponents for repression, and run roughshod over civil liberties protections. It’s bad enough that three of the world’s top scholars of comparative democracy — Steven Levitsky, Daniel Ziblatt, and Lucan Way — have concluded that the United States has crossed the line into a form of authoritarianism.

On the other hand, there is little indication that Trump has been able to create a lock on power — or even significantly compromise the fairness of elections. Democrats dominated elections in 2025, anti-government activists operate freely, and the media is (mostly) as independent and critical as it was before Inauguration Day. When I spoke to Levitsky in December, he told me that Trump was failing “at consolidating autocratic power.”

For this reason, my own view is that the United States is still best classified as democracy, albeit a much weakened one. V-DEM, the leading academic metric of democracy, distinguishes between two classes of democracy — the stronger liberal democracy and weaker electoral democracy. When V-Dem releases its ratings for the past year, I expect the United States will fall from the former into the latter.

However, my confidence is low. What’s happening in the US is unprecedented for the world’s hegemon, and there is at least some credible evidence of bias in global democracy ratings — making the ultimate outcome a bit tricky to say for sure. —Zack Beauchamp

Democrats will take back at least one house of Congress (95 percent)

If the last one was tricky, this one is straightforward. There are at least five clear reasons to believe Democrats are headed for a midterm romp.

Point 1: In modern American politics, the president’s party almost always performs poorly in midterms.

Point 2: The Democratic Party is increasingly strong with college-educated voters, who tend to turn out more reliably in midterms than non-college voters — meaning the party has a structural leg up in those contests.

Point 3: Trump is an especially unpopular incumbent. The only 21st-century president with equivalently bad numbers at this point in his term was Trump himself, who experienced a massive electoral wipeout in the 2018 midterms. And there is real evidence Trump’s coalition is fraying from the inside.

Point 4: Democrats have dominated 2025 elections so consistently that it has become a meaningful indication of 2026 performance.

Point 5: Voter dissatisfaction is driven by a combination of affordability and concerns about his extreme policies in areas like immigration, and the White House seems either unable or unwilling to change in response to these concerns.

For all these reasons, Democrats are basically a lock to take back the House — barring hard-to-pull-off election tampering or some kind of unforeseen event that transforms the political environment. The Senate map is unfavorable, making it a much tougher fight, but they’re still competitive given the fundamentals. —ZB

At least one major function remains at the Education Department (70 percent)

The dismantling of the Education Department was one of the biggest stories in the early days of Trump’s second term, as the administration fired hundreds of staffers and Education Secretary Linda McMahon promised to lead the department on its “historic final mission.”

The president can’t actually dissolve the department without an act of Congress, but his administration has been moving bits of it to other agencies since the spring. In November, the White House announced perhaps the biggest shift yet, moving programs supporting K-12 students to the Labor Department, with other functions parceled out to the Departments of Health and Human Services, Interior, and State.

However, experts have long warned that other departments don’t have the expertise to take over Education staffers’ work, and the moves that have already occurred have reportedly been plagued with problems. Now Republican lawmakers are starting to voice concerns about what happens if the administration tries to transfer special education programs to another department, a move it has not yet made but hasn’t ruled out.

The Trump administration has already done lasting damage to the department, experts say. But getting rid of an agency is a lot harder in practice than in theory, and with Republicans starting to throw up warning signs, it’s more likely than not that at least one function of the department will remain through the end of next year. —Anna North

The Supreme Court will rule against Trump in the tariffs cases currently before the Court (70 percent)

To date, at least three federal courts have ruled that President Donald Trump exceeded his power under the International Emergency Economic Powers Act (IEEPA), when he imposed a broad range of constantly shifting tariffs on foreign imports. The Supreme Court is likely to join these three courts before the close of its current term.

For the most part, this Supreme Court’s Republican supermajority has been extraordinarily loyal to Trump. This is, after all, the same Court that held that Trump may use the powers of the presidency to commit crimes. But the Republican justices do sometimes break with Trump on issues that divide Republicans, and especially on issues that divide conservative legal elites.

The tariffs cases are just such an issue. At least some of the lawsuits challenging the tariffs were brought by right-leaning legal shops that hew to the GOP’s more traditional, libertarian views on foreign trade. Numerous Republican luminaries have joined briefs opposing the tariffs,

Including former Sen. John Danforth (R-MO), an early mentor to Justice Clarence Thomas. Over the spring, at a conference hosted by the conservative Federalist Society, a number of speakers criticized the tariffs and questioned their legality.

At the Supreme Court argument on the tariffs in November, the Court’s Republicans did, indeed, appear divided on whether to back Trump. While some members of the Court defended the tariffs, Chief Justice John Roberts and Justices Neil Gorsuch and Amy Coney Barrett — all Republicans — asked very skeptical questions of Trump’s lawyer.

It is always dangerous business to predict that this Supreme Court will break with a Republican president, which is why I still think there is a 30 percent chance that Trump prevails. And even if Trump does lose this round of litigation, he is likely to attempt to reinstate at least some of his tariffs by invoking other statutes. But my prediction will come true if the Court rules that Trump exceeded his authority under the IEEPA when he imposed his tariffs on imports. —Ian Millhiser

Trump will replace at least one member of the Supreme Court by the end of 2026 (75 percent)

Trump is unpopular — a recent Associated Press poll pegs his approval rating at 36 percent — and his party just got hosed in the 2025 elections. Republicans are still favored to hold onto the Senate after the 2026 midterms, largely because the Senate is malapportioned to favor small states that tend to vote for the GOP, but the Republican Party is in a deep enough hole that it could lose both houses of Congress.

And if the Democrats do take the Senate, they can prevent Trump from ever confirming another federal judge again. Which brings us to 75-year-old Justice Samuel Alito.

Alito is the Court’s most unapologetic partisan. If you want a full rundown of Alito’s history of rulings favoring the Republican Party, I encourage you to read my profile of him entitled “The Republican Party’s man inside the Supreme Court.” The short of it is that he’s often willing to embrace arguments that even his fellow Republican justices find embarrassing, at least when those arguments favor the GOP or its preferred policy outcomes.

If Alito retires while Republicans still control the Senate, he can be confident that his replacement will be a Republican who shares his views on the overwhelming majority of issues. He might even be replaced by one of his former law clerks.

If Alito does not retire, by contrast, he risks losing his last chance to retire under a Republican president and a Republican Senate. In the worst case scenario (from Alito’s perspective), he could die after Democrats regain both the White House and the Senate, ensuring that he will be replaced by his ideological opposite.

There’s also a chance that a different justice could either retire or die. Thomas is 77. Justice Sonia Sotomayor is 71. Roberts is 70. If any justice leaves the Court in 2026, a Republican Senate will almost certainly confirm Trump’s nominee to replace them.

That said, there is a chance that Alito and his fellow Republican justices are enjoying the power that comes with being part of a six-justice supermajority so much that they won’t want to give it up. But Alito has been such a reliable partisan during his time on the bench that it would be surprising if he denied his party its best chance to replace him with a younger version of himself. —IM

The world

Benjamin Netanyahu will not be the prime minister of Israel by the end of the year (65 percent)

Netanyahu has led the Israeli government for 15 of the last 16 years. He has weathered indictments, a criminal trial, coalition fractures, and of course the horrors of the Gaza war. Why would anyone bet against him in the 2026 elections (currently scheduled for October)?

The answer, I think, is that he has been living on borrowed time since October 7, 2023.

After that day’s atrocities, Netanyahu’s poll numbers collapsed — with most Israelis blaming him and his government for Hamas’s successful attack. His survival since then has had nothing to do with voters, and everything to do with coalition management: He has managed to prevent his far-right coalition partners from defecting and triggering early elections. But in 2026, there will be elections — and all indications are that his coalition doesn’t have the votes.

“The Netanyahu government has not been able to win a majority in any credible survey,” Dahlia Scheindlin, a leading Israeli pollster, told me last year.

That said, you really do not want to count Netanyahu out. And there are easy-to-imagine scenarios where he survives despite his obvious problems.

Currently, the best-polling opposition party is led by former Prime Minister Naftali Bennett. While Bennett is strongly anti-Netanyahu, he is also a right-winger — and to form an anti-Netanyahu government, polls suggest he’d likely need support from a broad coalition, including the left and even an Arab party. You can easily imagine Bennett failing to overcome the opposition’s ideological divisions and striking some kind of deal with Netanyahu instead. Or you could imagine protracted coalition negotiations that leave Netanyahu in power for months after the October elections, even if he is deposed in 2027.

The point is that there’s a lot of uncertainty here. But I’m going to bet on the most consistent thing: Polls showing that a clear majority of Israelis are done with Bibi. —ZB

There will not be a ceasefire, agreed to by both Ukraine and Russia and observed for at least 30 days, by December 31, 2026 (60 percent)

A salon damaged by Russian bombing in Kostiantynivka, Ukraine, December 28, 2025.

| Diego Herrera Carcedo/Anadolu via Getty Images" data-portal-copyright="Diego Herrera Carcedo/Anadolu via Getty Images" />

The Trump administration has been pushing hard for a ceasefire deal in recent weeks and there was some optimism it might end before Christmas. But the underlying dynamics of the conflict are still the same and still make an end to the war in the coming months more unlikely.

Despite the heavy casualties Russia is taking, the damage to its economy inflicted by sanctions, and the slow pace of progress on the battlefield, Russian President Vladimir Putin believes he is winning the war and is unlikely to be satisfied with any deal that does not severely curtail Ukraine’s sovereignty. It’s not even clear if the 28-point plan cooked up by his envoys Steve Witkoff and Jared Kushner in November, which was heavily tilted toward Russian interests, would have been enough to satisfy him.

On the other side, Ukrainian leaders mostly accept at this point that they’re unlikely to regain all of the territory currently held by Russia by military force. But they are just as unlikely to accept Trump’s recent demands that they cede the so-called fortress belt of heavily defended positions in eastern Ukraine, something that would be suicidal fairly likely event that Russia restarts its war in a few years. And while NATO membership may be off the table at this point, Ukraine is likely to insist on security guarantees from NATO countries that will probably be unacceptable to the Russian side.

While US support for Ukraine gives it significant leverage, European countries are now the primary economic and military backers of the Ukrainian war effort and Ukrainians are making far more weapons of their own, including the ubiquitous drones that are playing such a vital role on the battlefield.

For all Trump’s public attacks on Ukraine, the United States is still providing intelligence support to the Ukrainian military and selling the country for weapons (in many cases, paid for by Europe). And if the past year’s back and forth is any indication, Trump’s current pro-Moscow tilt could shift.

Trump’s success with the Gaza ceasefire showed that these deals can come together much more quickly than many expect, but for a variety of reasons, the combatants in Ukraine are less susceptible to American pressure and less willing to call off the fighting. Most likely, Ukraine is facing a fifth year of devastating and brutal war. —Joshua Keating

Between January 1 and December 31, 2026, China does not impose a full blockade of Taiwan or launch a declared invasion (75 percent)

In 2021, Adm. Phil Davidson, then the head of Indo-Pacific Command, told Congress he believed China would likely seek to achieve its ambition of taking control of Taiwan “in the next six years.” We’re now approaching the later end of what has become known in defense circles as the “Davidson window.” But for the moment, war — or something close to it — still seems unlikely. The biggest question mark around a military scenario in Taiwan is whether the US would intervene directly to defend the island. And the best case for the argument that China will move soon is that President Donald Trump’s words and actions have given little reason to believe he would do that. But an amphibious invasion of a mountainous and densely populated island with a hostile population is still a daunting prospect even if the US doesn’t get involved.

A blockade or quarantine might be more likely, something Taiwan’s economy is vulnerable to, but the island’s importance to the global tech economy means the fallout from a blockade would be both massive and widespread. (One analysis predicted a blockade of Taiwan would cost the world $2 trillion in lost economic activity.) And the US is not the only country that might come to Taiwan’s aid: Japan’s new prime minister recently enraged Beijing by suggesting a Taiwan crisis would be a survival threatening situation for Japan, meaning it would have legal justification to deploy its military.

To put it bluntly, at the moment, Xi Jinping has a good thing going with Trump, who is seeking better trade relations with China and has even gone so far as to agree to sell advanced microchips that the Chinese never even asked for. China may also be holding out for the possibility of “peaceful reunification.” The island’s major opposition party, the Kuomintang, now favors much closer relations with Beijing.

We should absolutely expect more economic pressure on Taiwan and its supporters abroad, more moves to block diplomatic contacts between Taiwan and the outside world, more influence campaigns and propaganda directed at the Taiwanese public, and even possible “gray zone” attacks targeting Taiwan’s infrastructure, such as undersea communications cables.

Putin’s decision to invade Ukraine showed that sometimes autocratic leaders can make risky moves that seem to make little sense from the outside, but assuming Xi is a bit more level-headed, he’s unlikely to gamble it all on an invasion or blockade in the coming year. —JK

Economy

At least two more states will pass laws effectively ending apartment bans (single-family-only zoning) in most residential areas statewide (45 percent)

The last few years have seen the birth of a new paradigm in how housing in the United States is regulated and built. Ever since the widespread adoption of zoning codes over the last century, it’s been local governments — cities, suburbs, small towns — that decide what’s allowed to be built, usually to an extreme degree of prescriptiveness. Most residential land across the country is zoned exclusively for detached single-family homes — no duplexes, triplexes, or apartment buildings allowed. That is, as I wrote about last year, what’s fundamentally at the root of the great American housing shortage and housing affordability crisis.

But those rules are steadily, if slowly and unevenly, starting to change. Many states have passed legislation that begins to unwind the morass of local obstacles to building homes, with single-family-exclusive zoning being a frequent target. While this trend is technically a form of centralization, I think it’s better to think of it as a kind of deregulation that gives power back to people to create things in their communities. California, Maine, Montana, Oregon, Vermont, and Washington all now have laws requiring local governments to allow at least duplexes, and in some cases even more homes, on lots zoned for single-family homes in many residential areas. Several other states are considering similar bills, and more will probably be introduced this year.

These state-level zoning preemption laws are, in reality, usually enormously complex and often include carveouts and exceptions that were needed to get the legislation over the finish line because local opposition to new housing can be fierce. So while I think we’re extremely likely to see more states pass housing liberalization laws in 2026, I think the chances that two more states pass laws with my exact criteria — ending single-family zoning in the residential areas that cover most of the state’s population — are just under 50-50. —Marina Bolotnikova

Total billionaire wealth will exceed $17 trillion, as calculated by the UBS Billionaire Ambitions report (85 percent)

The grass is green. The sky is blue. The rich get richer.

Some things are just common sense. But actually, the wealth of the very wealthiest people does not always get bigger year after year. Take 2022, for example, when stock market woes made the world’s billionaires about $2 trillion poorer than they were the year before. Womp, womp.

But ever since then, the ultra-rich have indeed only gotten richer. A new billionaire was born every 37 hours of 2025, lifting the total number of billionaires to nearly 3,000 and their collective wealth to a record-shattering $15.8 trillion, according to the UBS Billionaire Ambitions report. Many have gotten rich off the AI boom, while others are heirs and heiresses, whose inheritances grew by a collective $297.8 billion last year as part of a giant wealth transfer that’s just getting started.

As long as nobody bursts their bubble, the ultra-rich will probably just get richer in 2026. And if their wealth keeps growing at the rate it has been, they’ll very likely be sitting on over $17 trillion by the time UBS publishes its report next winter. —Sara Herschander

The US will experience a recession in 2026 (55 percent)

Even in “good” times, the US economy is a balancing act between consumer spending, business investment, financial conditions, and policy choices.

| Spencer Platt/Getty Images" data-portal-copyright="Spencer Platt/Getty Images" />

Recession forecasts are the meteorology of economics: Everyone complains when you’re wrong, and nobody sends thank you notes when you’re right. Still, the reason I’m slightly over 50 percent is simple: Late-cycle economic risk is real, and the list of plausible triggers — the AI bubble popping, trade policy finally hitting home, a major international crisis — is long.

Even in “good” times, the US economy is a balancing act between consumer spending, business investment, financial conditions, and whatever policy choices Washington makes in a given week. It doesn’t take a Great Depression-level shock to tip that balance — sometimes it’s just interest rates staying tighter longer than expected, a confidence shock, or a geopolitical event that hits energy and trade. And if 2020 taught us anything, it’s that the economy can fall down the stairs faster than it can climb them.

For scoring, I’d define “recession” as a National Bureau of Economic Research-dated recession that begins in calendar year 2026. If the NBER hasn’t ruled by the time we do our year-end grading (they are not known for sprinting), we’ll use a proxy: two consecutive quarters of negative real GDP growth in 2026.

Why only 55 percent? Because the US has a stubborn capacity to muddle through — until it doesn’t. —BW

Animals

The share of cage-free eggs in the US will not surpass 50 percent in 2026 (60 percent)

At the end of 2024, cage-free eggs accounted for 38.7 percent of the US egg supply. By September 2025 — the most recent data available — that figure hit 45.3 percent. It was a major shift for such a short period, and equates to millions of egg-laying hens no longer spending their entire lives in tiny cages.

I think that in 2026, this trend will continue, but not fast enough for the US egg supply to reach 50 percent cage-free by the end of September. And that’s because a few big events occurred in 2025 that spurred this momentum that won’t occur next year.

The first is that laws in three states — Arizona, Colorado, and Michigan — that require eggs sold to be cage-free went into effect in 2025 (though Arizona quickly delayed its implementation by years). No new laws will go into effect next year.

Additionally, over the last decade, hundreds of food companies pledged to source cage-free eggs, and many set a 2025 deadline. While a lot of them have not followed through on their pledge, a lot inched closer during this deadline year.

I could be — and hope I will be — wrong, and there are two reasons why I might lose this prediction. The first is that animal advocacy groups are now focused on pressuring grocery chains to meet their cage-free pledges, and if they’re successful in 2026, that could quickly tip the scales, since grocery stores account for where most eggs are sold. Second, there’s bird flu — if the virus were to disproportionately hit cage farms this winter and spring, that would affect the ratio of cage-free to cage eggs for much of 2026.

The food industry’s rapid move away from cages for egg-laying hens is a major success story for the modern animal rights movement, and hitting 50 percent of the US egg supply will be an important milestone. I think it’ll happen soon — let’s say by March 1, 2027 — but I don’t think it’s in the cards by the end of September, 2026. —Kenny Torrella

At least one US state will ban lab-grown meat in 2026 (60 percent)

In 2024, Florida and Alabama banned the production and sale of lab-grown, or cell-cultivated, meat. They represented unabashed protectionism — favoring livestock farmers over startup food companies — and hollow, conspiratorial culture war posturing (when he signed the bill into law, Florida Gov. Ron DeSantis implied this was a contest between real Floridians and globalist elites).

Even though many ranchers and farming groups have opposed the bans, arguing that it looks bad to outlaw your competition, five more states passed similar laws in 2025 — three with full-on bans (Mississippi, Montana, and Nebraska), and two with two-year bans (Texas and Indiana). In many other states, lawmakers introduced similar bills that failed, and I figure at least one will succeed next year (for the purpose of accuracy, I’ll count a temporary ban as a ban).

I’m not particularly confident, however, because some states seemed to have settled on strict, unfavorable labeling requirements for cell-cultivated meat producers as opposed to banning the product altogether. And in some states, the bills have proven controversial (for example, many ranchers in Wyoming were opposed to a ban on libertarian grounds).

At this moment, the bans mean little in practical terms — only a few restaurants around the country serve cell-cultivated meat, and in small quantities. But the bans could pose a problem for the industry down the road if they figure out how to affordably produce cell-cultivated meat at scale. —KT

The US will authorize mass bird flu vaccination for at least one major US poultry category — egg-laying hens, broiler chickens, or turkeys (35 percent)

The US is entering its fifth year of a truly ghastly bird flu outbreak. It’s caused dozens of human bird flu cases across the country, it’s sparked an outbreak in dairy cows, it’s sent egg prices soaring, and it’s been catastrophic for the tens of millions of chickens and turkeys who’ve died horrible deaths on infected farms. And all this is happening despite the fact that we already have vaccines that could dramatically blunt the damage.

So why, four years into this outbreak, have we managed to do so little to get avian flu under control?

It has more to do with bureaucracy and economic interests than scientific capacity. The American chicken meat industry exports a significant share of its product abroad, and the fear is that our trading partners would reject US chicken because of the challenge of determining whether a poultry bird is infected with avian flu or simply has antibodies from vaccination. So instead of vaccinating, the US has resorted to mass killing chickens and turkeys — quite painfully — in a mostly unsuccessful attempt to control the spread.

As the outbreak stretches on, and egg and turkey producers complain that they aren’t allowed to vaccinate because of the chicken industry’s trade concerns, pressure has mounted for US regulators to approve a plan to start vaccinating poultry birds across the country — something that ought to be a no-brainer given that, as Vox’s Kenny Torrella has pointed out, the costs of managing the outbreak have been much higher than the value of the chicken industry’s exports. As of last summer, the US Department of Agriculture was reportedly working on such a plan.

Will we start routinely vaccinating in 2026? We’re closer than we have been in any previous year, but securing assurances from trade partners is hard, long work, as is devising a plan for vaccine rollout that satisfies those partners, and all indications are that we’re not close yet. If we start to see more severe bird flu spread in 2026 and sustained spikes in egg prices, the USDA’s calculus might change. But for now, I think we’re less likely than not to see the agency authorize vaccination as part of a standard avian flu control program in poultry birds (rather than just as part of limited pilots or experimental uses). —MB

Climate

Global data center electricity demand will stay below 3 percent of total electricity in 2026 (80 percent)

A sign on a rural Michigan road opposes a planned $7 billion data center on southeast Michigan farm land. Opponents say the Data Center could raise residential electricity rates and endanger the water supply.

Jim West/UCG/Universal Images Group via Getty Images

" data-portal-copyright="

Jim West/UCG/Universal Images Group via Getty Images

" />

Per the International Energy Agency, data centers consumed 1.5 percent of the world’s electricity in 2024, around 415 terawatt-hours. Though these massive, energy-hungry facilities are proliferating at a rapid pace, they’re still a small fraction of humanity’s energy use.

Tech companies say they need many more of them, particularly to run their AI products, but data centers have an image problem. They are starting to wear out their welcome in some communities and are being thoroughly shunned in others. Only 44 percent of Americans say they would want one of these giant humming boxes near them. Speculation around their energy demand is already starting to raise electricity prices for consumers in some markets.

Now some environmental groups and activists are already calling for a moratorium on new data center construction, not just voting down individual projects, and at least one community has officially imposed a pause.

There are also strains on the global supply chain for data center components, so even places ready to go on a construction spree will have to wait for parts to catch up. Additionally, more power generators are continuing to come online, so the percentage share that goes to data centers won’t rise as quickly. —Umair Irfan

At least one Category 5 hurricane makes landfall in the continental US, as defined by the National Hurricane Center (10 percent)

The United States lucked out in 2025 with no major hurricane hitting the mainland. However, it’s only a matter of time before one does so again. The question is how strong it will be. There are typically 14 storms strong enough to be given a name in any year, but only 45 were ever known to have reached Category 5 strength, with sustained winds at 158 miles per hour. Fewer still maintained their full strength as they reached the shore.

The last Category 5 hurricane to hammer the continental US was Hurricane Michael in 2018, so baseline chances of this happening again next year are fairly low. The year 2026 is poised to start as a La Niña year, where the surface of the Pacific Ocean cools to below-average temperatures. That tends to create more favorable conditions for hurricanes in the Atlantic Ocean. The pattern is then lonely to shift into a neutral phase that has minimal effects on cyclones in the Atlantic. The other key variable is how much heat is in the Atlantic Ocean. Hurricanes run on warm water, and the fever of record-high temperatures broke in the Atlantic Ocean this year.

Add to that the variability in how hurricanes travel and you have a fairly low chance of the most powerful type of hurricane hitting the continental US at maximum power next year. —UI

Science and technology

At least one state-of-the-art AI system can complete a task that takes humans 16 hours, succeeding on at least half of its attempts (75 percent)

One of the past year’s most striking AI-related visuals was a graph showing that the length of tasks AI can do is doubling every seven months. This may seem a bit in the weeds, but it’s actually really important, because it speaks to AI’s growing ability to work autonomously. According to METR, the research group that made this graph, Claude Opus 4.5 has already hit four hours and 49 minutes, which means that the chatbot is expected to succeed at least 50 percent of the time on tasks that took humans that long. Extrapolating from this graph, I predict that at least one AI model will hit at least 16 hours by the end of 2026. I’m making this prediction with 75 percent confidence.

I could go higher, but I won’t, because a few variables could still change the trajectory. For example, if compute growth slows, we could see substantial delays in capability milestones. I also want to emphasize that you shouldn’t take this to mean that AI will put you out of work by the end of 2026: What’s being measured here is AI’s ability to succeed at very particular tasks, not its ability to generalize to the whole of what you can do. — Sigal Samuel

Congress will not pass, and Trump will not sign, any comprehensive federal legislation primarily focused on AI safety (90 percent)

The White House has come out strongly against state-level AI regulation, releasing an executive order in December saying that the “Administration must act with the Congress to ensure that there is a minimally burdensome national standard — not 50 discordant State ones.”

But it’s very unlikely that we’ll see comprehensive federal legislation in 2026 requiring AI companies to implement safety plans. For one thing, there is no consensus on what shape such a national framework should take. For another, the White House’s attempt to ban state-level regulation (with the idea of putting in a national framework instead) has proven extremely unpopular, including among Republicans. Plus, with so much tech lobbying aimed at relaxing regulation rather than entrenching it, there’s little incentive for the White House to push through comprehensive federal legislation on safety.

Taken together, all this leads me to think that while Congress may pass more specific AI provisions in 2026 (for example, related to national defense), it won’t pass a comprehensive national standard when it comes to actually keeping us safe from AI. —SS

At least one primarily AI-generated song reaches No. 1 on the Billboard Hot 100 chart (60 percent)

This is the kind of prediction that sounds silly right up until it’s not. Wholly AI-generated music has already crossed one major threshold, when the country track “Walk My Walk,” by the AI band Breaking Rust, topped Billboard’s Country Digital Song Sales chart. One survey found that 60 million people were using AI tools to make music, while the streaming platform Deezer reported that a third of the tracks uploaded each day were AI generated.

The remaining barrier to AI music colonizing your ears isn’t capability so much as distribution: You don’t hit No. 1 because you made a great song — you hit No. 1 because the machinery of attention (TikTok, streaming playlists, fandoms, and labels) decides to make your song unavoidable. And I could see the sheer novelty factor pushing at least one AI generated song to the top of the pops.

So what counts as “primarily AI-generated” here? For scoring purposes, I’d define it narrowly: The core musical content (melody/arrangement and a substantial share of the vocals or instrumentation) must be generated by an AI system, and that fact has to be publicly acknowledged by the creators or credibly reported: “AI was used in mastering” or “a producer used AI for a synth patch” — aka AI as a means to supplement human-made work doesn’t count. If it’s essentially an AI-made track with human polishing, it qualifies.

Why 60 percent? Because the incentives of novelty, speed and cost all line up. The big uncertainty is backlash: legal, cultural, or platform-level. But history suggests that if something can go viral, it eventually will. —BW

Elon Musk will exit the Giving Pledge (55 percent)

Musk is on track to become history’s first trillionaire. His fortune is already so gargantuan that if he wanted to, he could end world hunger and subsidize a free national preschool program and still have hundreds of billions of dollars to spare.

But don’t bet on it, because the world’s richest man may soon become the first person ever to go take-backsies on the Giving Pledge, a promise by the ultra-wealthy to donate half of their wealth in their lifetime or upon their death.

To be fair, plenty of other signatories have quietly died without fulfilling their pledge. But Musk has also drifted far away politically from who he was when he signed in 2012, and his qualms about philanthropy — including that of his fellow pledgers — are no secret. He thinks it is “extremely difficult” to give money well. MacKenzie Scott is “concerning.” Nonprofits are “money laundering” schemes. Philanthropy is “bullshit.” And the pledge’s founder Bill Gates, Musk told his biographer Walter Isaacson, is “categorically insane (and an asshole to the core).”

Oh, and his good friend Peter Thiel has been openly encouraging Musk — whose charitable foundation has regularly failed to meet the minimum legal giving requirements anyway — to unsign. Altogether, it’s become more likely than not that Musk will publicly bow out of the Giving Pledge before December 31, 2026. It could come in the form of a quiet delisting on GivingPledge.org, but chances are we’ll find out on X before anywhere else. —SH

There will be a satellite collision in low Earth orbit (75 percent)

Space is getting awfully crowded.

About 15,000 satellites currently orbit Earth. That number has risen exponentially in recent years due to megaconstellations, large satellite networks launched by private companies like SpaceX and Amazon to provide broadband internet access around the world. Most of these satellites are in low earth orbit (LEO), or 1,200 miles or less above the planet’s surface. As of late October, there were at least 12,000 active satellites in LEO — and just over 66 percent are a part of SpaceX’s Starlink constellation, which aims to eventually have up to 42,000 satellites.

We’ve launched more satellites to LEO in the last four years than we have in the previous 70 years combined. By 2040, we should expect to see more than 560,000 satellites in orbit based on planned launches. It’s hard to predict exactly how many satellites we’ll have by the end of 2026, but we know that Starlink and other megaconstellations will continue to grow.

The more satellites we have, the greater the chance that they will collide into one another or “space junk” — debris from human-made objects like defunct satellites, bits of spacecrafts, and old rocket parts. Various countries have space traffic management systems to protect against this, but they certainly aren’t fail-safe, especially given the rate at which new satellites are being launched into orbit and the increasing risk of collisions that comes with that.

On December 9, a Starlink satellite narrowly avoided colliding with a Chinese satellite. Space X claimed that the Chinese operator didn’t share its location data. Starlink satellites can automatically change course to avoid objects, but they have to know they’re there for this to work. In the first half of 2025, Starlink performed more than 144,000 avoidance maneuvers. So yes, collisions are inevitable — they’re just a question of when. I’d say 2026. —Shayna Korol

Health

The US will approve at least one fully synthetic, small-molecule oral GLP-1 receptor agonist for obesity and/or Type 2 diabetes treatment (70 percent)

GLP-1 drugs like Ozempic and Wegovy are all the rage, but high demand has meant serious shortages. That’s partly because these drugs are complex peptides grown from living cells, a process that’s hard to scale. But that won’t be the case for long.

Eli Lilly, an American pharmaceutical company, has developed an oral GLP-1 pill that works like the injections but is structurally very different, more similar to an aspirin. A pill like that would be much cheaper, won’t require cold storage, and can be pressed into pills by the billions. In pivotal trials, the drug showed weight loss rivaling the injections.

Lilly is submitting for FDA approval by year’s end, and the drug has been selected for the FDA’s new priority voucher program, which can cut review times from 10 months to as little as two. The government has already struck a deal with Lilly capping Medicare patients’ costs at $50 a month if approved. And CEO Dave Ricks told CNBC he expects a global launch “this time next year.”

Already in late December, the FDA approved a pill version of Novo Nordisk’s Wegovy. But that’s still a semaglutide, or peptide — not what I’m covering with this prediction. But if Lily’s approval goes through, we’ll be in a true era of GLP-1 abundance. —Pratik Pawar

Robert F. Kennedy Jr. is still serving as health secretary by the end of the year (60 percent)

RFK Jr. has never been a natural fit for the Trump administration. A longtime Democrat with a history of environmental advocacy, he was initially useful to Trump largely because he brought in voters supportive of his Make America Healthy Again movement. But his anti-vaccine advocacy has gotten him into trouble with Republican senators and occasionally put him out of step with Trump, who said in September that “you have some vaccines that are very amazing.”

Indeed, if I’d been making this prediction in the fall, I might have given Kennedy less than even odds of staying in his position through 2026. However, he has scored wins lately, like rolling back the federal recommendation that infants receive the hepatitis B vaccine at birth (to be clear, experts say getting rid of the recommendation is dangerous and could lead to unnecessary deaths). He has also managed to avoid real political fallout around the release of Olivia Nuzzi’s memoir about their alleged affair. Vaccine skeptics are reportedly excited about their recent victories and looking forward to more.

Neither vaccine opposition nor MAHA more generally are truly core to Trump’s governing project, to the extent that he has one, and it’s possible to imagine a post-midterm shakeup of the US Department of Health and Human Services. For now, however, the odds favor Kennedy keeping his job. —AN

The World Health Organization will officially withdraw the United States’s measles elimination status (75 percent)

The United States earned measles elimination status from the WHO in 2000, after decades of a successful vaccination campaign. More than 90 percent of children received the measles vaccine — and Americans widely agreed on its value. In the following years, with rare exceptions, the only cases in the US were brought here from other parts of the world where measles was still more widespread.

Not anymore.

Measles vaccination rates have been sliding for years, and 2025 brought the biggest single outbreak in more than three decades, seeded in West Texas among a religious community that is skeptical of vaccinations. Even as that outbreak petered out over the summer, after more than 700 cases and three deaths, local outbreaks have persisted in Utah and South Carolina.

The WHO’s criteria for revoking measles elimination status is 12 months of continuous transmission. Considering the same strain of the measles virus that was present in Texas in January was still circulating as of November, it doesn’t look good.

It seems to me that only a massive effort from the federal government could stamp out the disease in time — but that appears far less likely than the Kennedy-led health department limiting access to the measles vaccine next year. Instead, it looks like a pretty safe bet that one of the most contagious viruses known to humanity will continue spreading long enough to undo one of the US’s signature public health wins. —Dylan Scott

Culture

Beyoncé will release a rock album (55 percent)

Will Beyoncé release a rock album in 2026?

| Emma McIntyre/Getty Images for the Recording Academy" data-portal-copyright="Emma McIntyre/Getty Images for the Recording Academy" />

Not many of my colleagues know this about me, but I’m a huge Beyoncé fan — and how could one not be? She has a voice like honeyed velvet, she can belt like no one else alive, and she can tear through choreo in six-inch heels like she’s just getting warmed up. Her creative instincts have made her one of America’s most consistently admired stars for over two decades, and she’s nothing if not incredibly versatile.

It’s already widely speculated that the third album in Beyoncé’s Renaissance trilogy (the first two being 2022’s Renaissance and 2024’s Cowboy Carter) will be rock ’n’ roll-adjacent, with many reports citing the rock songs she’s already released on Lemonade and her most recent album, plus the numerous rock-coded Easter eggs she’s been dropping over the last year. But she’s also been manifesting a bigger rock project ever since her jaw-dropping backbend set to electric guitar at a 2009 performance of “Freakum Dress,” and probably for even longer.

Her Renaissance trilogy, so far, has explored the Black musical roots of modern pop music, with each release encompassing not a fixed genre but a sonic world with porous borders. So while rock is a narratively satisfying guess for Beyoncé’s next act, there’s also a great deal of uncertainty — she’s rarely straightforward or predictable.

Nevertheless, I’ll place my bets that she’ll have an album out this year with rock or a rock subfield as its primary genre, as defined by at least one major music chart or streaming platform (Billboard, Apple Music, or AllMusic), or as defined by album reviews in a majority of the following outlets: Pitchfork, Rolling Stone, the New York Times, The Guardian, and Vulture. A tad overconfident? Perhaps. But we could all use a hard cultural pivot from the last few years of country music and aesthetics, and I can’t wait to see what Beyoncé will do as rock frontwoman. —MB

Jacob Elordi will be nominated for an Oscar for Best Supporting Actor for his portrayal of the creature in Frankenstein (70 percent)

Oh, Mr. Kissing Booth. I didn’t think you had it in you, but your sorrowful, baby doe eyes as the creature has endeared me!

I went to see Frankenstein in IMAX with one of my friends, and I knew that I was going to walk into a monster-sympathetic adaptation. (It’s Guillermo del Toro we’re talking about, he of Pan’s Labyrinth and The Shape of Water.) I’m a big fan of the book, and was eager to see how Elordi would interpret the creature’s curiosity, rage, and desire for love. Elordi’s creature was more than I could have ever hoped for. Elegant, childlike, and grotesque, all wrapped into one lanky 6-foot-6-inch body — a beautiful foil to Oscar Isaac’s impetuous Victor. I entirely forgot this is Nate from Euphoria! And apparently so did everyone at Cannes.

He will be nominated for Best Supporting Actor, but the odds he wins are lower, depending on who from One Battle After Another is nominated, either Benicio del Toro or Sean Penn. If it’s both, Elordi is cooked. —Izzie Ramirez

Inside Trump’s “no data, just vibes” approach to science

2026-01-26T11:00:13-05:00

Just trust him, right? | Julia Nikhinson/Pool/AFP via Getty Images

One of the biggest changes so far during President Donald Trump’s second term has been the steady degradation of basic data collection.

In some cases, moves have been driven by his ideological resistance to the numbers themselves; in others, by a desire to bury uncomfortable trends. And in many places, it’s simply the result of deep job and budget cuts that have left agencies unable to track the country they’re meant to govern.

Key takeaways

• The federal government is a key collector of vital data about the makeup of the country.

• President Donald Trump has long been hostile to data that contradicts his messaging and has presided over major rollbacks to data collection relating to the environment, public health, employment, demographics, and the weather.

• With less robust and accurate data, advances in science will slow down, Americans will have a murkier picture of the economy, and officials could miss important health trends. It will also further erode trust in public institutions.

Gathering basic data about the country is one of the key responsibilities of the federal government. After all, the census is mandated by the Constitution. Getting correct numbers about people, their health, the environment, and the economy is essential for taking an accurate snapshot of the country. These data are also the essential foundation for allocating resources and for sorting what works from what doesn’t.

Good numbers are a key accountability tool, and with the absence of data or lower-quality numbers driving decisions, it will be easier for leaders to mislead. Strip away the measurements and tallies, and the consequences pile up fast: Scientific research slows, early warnings about health threats get missed, economic policies become more volatile, and trust in institutions erodes even further.

Of course, good information can often have huge political consequences, which creates a strong temptation to fudge the figures.

But the Trump administration has gone far beyond its predecessors, cutting entire data-collection programs while putting ideologues in charge of fact-finding — all while pressuring agencies to support preordained conclusions. And if the White House has its way, even more rollbacks are in store.

Here are some of the most significant ways in which the White House has diminished our capacity to count and measure the country, and the world, this year.

1) Scaling back vital health surveys

Over the spring, the Trump administration laid off federal workers responsible for collecting basic information about people’s well-being and put in motion the process to overhaul federal surveys to eliminate the questions related to racial minorities and LGBTQ+ people.

We may not think of the federal government as one of the most important pollsters in the world, but it is: The best data we have about everything from teen smoking to increases in obesity rates to how many people have health insurance has come from the government.

Among the estimated 3,000 employees laid off from the Centers for Disease Control and Prevention were staffers who conduct surveys assessing everything from pregnancy risks to youth smoking to sexual violence. Without that data, the country will be flying blind when new health trends emerge. And as the administration moved to erase certain underrepresented communities from data collection, it will be harder to know whether depression or anxiety are particularly high among LGBTQ+ people or whether certain populations are becoming more susceptible to hypertension or diabetes.

The White House justified the cuts partly in the name of reducing government waste and partly as part of its ongoing crusade to erase any protections for and recognition of transgender or gender non-conforming people.

But that comes at a cost. The raw data that allows us to intervene and stop health problems are evaporating. —Dylan Scott, Vox senior health correspondent

2) Clawing back research grants

The National Institutes of Health, which awards upward of $40 billion in grants to scientific researchers every year, is the single biggest funder of independent scientific inquiry in the world.

But this year, the administration slashed its financial support for those research projects by an estimated $2.7 billion while proposing billions more in future cuts — cutting off another vital source of information about what’s driving changes in the population’s health and how any emerging problems might be fixed.

The list of canceled NIH projects, as documented by ProPublica, is long and varied. Scientists have been working for years to diversify their clinical trial participants, to collect better data that better reflects the wider population. One such project, to improve the recruitment for Alzheimer’s disease clinical trials, was being funded by an NIH grant — and it was cut by the Trump administration. Another grant uncovering new data on how contaminated drinking water affects fetal development — cut. New research into how discrimination affects the mental health of young Hispanic people, into the maternal health of Black women, into the driver of the disproportionate death rate from cervical cancer among Black women — cut, cut, and cut.

These are the kinds of nuanced scientific questions that the federal government’s surveys can’t answer on their own. That’s why the US has long provided support to independent researchers who can provide us with answers. This system has relied on the trust of the scientific process.

But not anymore. —Dylan Scott, Vox senior health correspondent

3) Overhauling the childhood vaccination schedule

The administration has been busy overhauling the childhood vaccination schedule — based not so much on new facts but out of the deeply felt convictions of Health Secretary Robert F. Kennedy Jr. and his handpicked panel of vaccine-skeptical advisers that something must change given the declining public trust in vaccines.

For example, Kennedy’s vaccine advisers justified their decision to end the recommendation for a birth dose of the hepatitis B vaccine based in part on dubious data that they said suggested the vaccine’s immunity waned over time. But even some of the advisory committee’s own members, who were overruled on the final decision at a December meeting, questioned the evidence for the change.

“There is not a single case who is otherwise healthy who received the recommended schedule, of anyone who developed disease or is symptomatic or has chronic disease,” Cody Meissner, a Tufts University pediatrician and infectious disease specialist, said during the adviser meeting. “The evidence is very strong that there is lifelong immunity to hepatitis B.”

The new recommendations they approved did suggest, however, that if your child does not receive the birth dose, you should wait until they are at least two months old before giving it to them. At least two members of the committee argued that there was no scientific basis for the two-month recommendation, and no data had been presented to justify it.

“It’s unconscionable,” Hibblen said shortly before the final vote. Nevertheless, the change was approved.

As Wilbur Chen, an infectious disease physician at the University of Maryland, put it to me after watching the meeting: It calls to mind a magician with a sleight of hand. They were picking data, whatever it is that supports their argument.” —Dylan Scott, Vox senior health correspondent

4) Deleting climate change references

From the outset, the Trump administration has had federal climate change research in its crosshairs. Project 2025, the Heritage Foundation’s proposed agenda for Trump’s second term, said that the White House needs to “eradicate climate change references from absolutely everywhere.”

Trump has done much more than delete the words “climate change” though; his administration has taken climate-related tools, data, and reports offline.

The budget and staff cuts at agencies like NOAA — the main department monitoring weather and climate — have reduced data collection activities like weather balloon launches that are important for forecasting models. There have also been budget and personnel cuts to divisions that do key tasks for research and predictions like flying aircraft into hurricanes. The agency also retired its database of billion-dollar disasters, which had tracked the costliest extreme weather events across the country going back more than 40 years.

More recently, the US has withdrawn its last research ship from Antarctica, a key field site for climate research. And now Trump wants to dissolve the National Center for Atmospheric Research, an internationally renowned institution that White House budget director and Project 2025 author Russ Vought called “one of the largest sources of climate alarmism in the country.”

Climate research is about much more than understanding climate change; it’s a critical field for tracking evolving risks in the environment and threats to the economy. The federal government’s climate research work has long led the world, and its efforts will be hard to duplicate elsewhere. —Umair Irfan, climate correspondent

5) Handcuffing the EPA

The Environmental Protection Agency has a mandate to protect human health and the environment, but the Trump administration has been celebrating its efforts to constrain it. One of its strategies is to roll back efforts to monitor pollution and enforce regulations. For example, the One Big Beautiful Bill Act zeroed out fines for car manufacturers that violate vehicle fuel economy and pollution rules.

The EPA made it easier for industries to apply for exemptions to air pollution standards. The agency also scrapped grants for measuring pollutants in communities with industrial facilities.

This year, the EPA initiated the fewest lawsuits against polluters in 25 years. The Department of Justice’s environmental enforcement division, which handles EPA’s litigation, now has around half the number of lawyers it did at the start of the year. With declining enforcement, the government has fewer resources to monitor violations of pollution regulations, while industries face less pressure to track and reduce their impact on the environment.

The agency is cutting back on its scientific efforts as well. The EPA’s Office of Research and Development, which provides the scientific basis for its regulations for things like toxic chemicals and water contaminants, was shuttered over the summer. This includes the Human Studies Facility, one of the largest laboratories in the country, which studied how smog, smoke, and soot affect the human body.

Looking ahead, the White House wants to shut down existing satellites that track carbon dioxide and remove pollution monitoring capabilities from the next generation of weather satellites. And the EPA wants to end greenhouse gas reporting for major industrial polluters, which includes more than 8,000 facilities.

All of this means less visibility and accountability for the things that make our planet less livable. —Umair Irfan, climate correspondent

6) Assault on jobs data

If the Trump administration had had its way earlier this year, then we might not have known just how bad the job market is right now, with the unemployment rate now at its highest level in four years.

Over the summer, Trump — who has a history of rattling wildly inaccurate unemployment numbers — fired the head of the Bureau of Labor Statistics after the agency released revised jobs data that made the economy look bad. Trump’s first pick for a replacement, the chief economist of the Heritage Foundation, floated suspending the monthly jobs report altogether before bipartisan condemnation forced the White House to withdraw his nomination. The agency is currently being led by acting commissioner William Watrowski, a longtime civil servant, pending a new nominee from the Trump administration.

So, for now, the jobs data appears safe. But with about one-third of leadership roles at the agency vacant, and a president still very much in denial about how the numbers work, it’s unclear if they will stay that way for long. —Sara Herschander, Future Perfect fellow

7) Trump takes aim at quarterly earnings reports

Since 1970, American companies have been required to report their earnings on a quarterly basis — a cache of data offering transparency about public companies that is considered to be among the most reliable in the world.

But Trump would like to change that.

In a September Truth Social post, he advocated for the Securities and Exchange Commission to make firms report on a semiannual rather than quarterly basis. This would, according to Trump, “save money and allow managers to focus on properly running their companies.”

He tried to push this through during his first term, although nothing materialized then. But now the SEC is actively looking into this, and if Trump is successful, this would put the US more in line with UK and EU practices. But many companies in the American market are growing significantly faster than their European counterparts, and investors benefit from more — not less — frequent information. —Shayna Korol, Future Perfect fellow

8) Shaking up the Census

The centuries-old census is a very big deal. Its results can redraw voting districts and control the fate of hundreds of billions of dollars of federal funds for schools, roads, and hospitals.

That’s why it’s so important that the Census Bureau, the country’s largest statistical agency, gets its counting right. Exactly how the census asks Americans about themselves has evolved dramatically over the decades. During President Joe Biden’s term, the administration required the 2030 Census to include, for the first time, new checkboxes for “Middle Eastern or North African” and “Hispanic or Latino” participants under a question about race and ethnicity. This is a crucial change because with more accurate data for those previously undercounted populations, the country will be able to more effectively allocate resources and enforce civil rights legislation.

Unless, of course, the Trump administration gets to it first. A White House official recently said that the administration is considering revoking those changes — which were made to better capture people’s racial identities — amid a broader war against anything even remotely tied to diversity, equity, or inclusion.

Trump has also repeatedly attempted to exclude undocumented people from the census, which would be an unprecedented change. If either of those things happen, the country will likely be one step further away from understanding itself — and undercounted American communities will suffer the consequences. —Sara Herschander, Future Perfect fellow

A race to save what’s left

This, of course, isn’t Trump’s first time in office, nor is it his first attempt to manipulate, ignore, or erase the numbers. And researchers, nonprofits, and activists have raised the alarm before about losing access to quality government data.

There are now multiple groups working to rescue and archive federal statistics and websites, as well as guides for finding information that has gone missing.

But there’s only so much companies, universities, and NGOs can do to match the US government’s data-gathering scale and depth. A concerted effort from the White House to diminish or manipulate the numbers behind policies will be hard to counteract, and the effects will linger for years to come.

Will AI make research on humans…less human?

2025-12-12T09:30:09-05:00

If you’re a human, there’s a very good chance you’ve been involved in human subjects research.

Maybe you’ve participated in a clinical trial, completed a survey about your health habits, or took part in a graduate student’s experiment for $20 when you were in college. Or maybe you’ve conducted research yourself as a student or professional.

Key takeaways

AI is changing the way people conduct research on humans, but our regulatory frameworks to protect human subjects haven’t kept pace.
AI has the potential to improve health care and make research more efficient, but only if it’s built responsibly with appropriate oversight.
Our data is being used in ways we may not know about or consent to, and underrepresented populations bear the greatest burden of risk.

As the name suggests, human subjects research (HSR) is research on human subjects. Federal regulations define it as research involving a living person that requires interacting with them to obtain information or biological samples. It also encompasses research that “obtains, uses, studies, analyzes, or generates” private information or biospecimens that could be used to identify the subject. It falls into two major buckets: social-behavioral-educational and biomedical.

If you want to conduct human subjects research, you have to seek Institutional Review Board (IRB) approval. IRBs are research ethics committees designed to protect human subjects, and any institution conducting federally funded research must have them.

We didn’t always have protection for human subjects in research. The 20th century was rife with horrific research abuses. Public backlash to the declassification of the Tuskegee Syphilis Study in 1972, in part, led to the publication of the Belmont Report in 1979, which established a few ethical principles to govern HSR: respect for people’s autonomy, minimizing potential harms and maximizing benefits, and distributing the risks and rewards of the research fairly. This became the foundation for the federal policy for human subjects protection, known as the Common Rule, which regulates IRBs.

Men included in a syphilis study stand for a photo in Alabama. For 40 years starting in 1932, medical workers in the segregated South withheld treatment for Black men who were unaware they had syphilis, so doctors could track the ravages of the illness and dissect their bodies afterward.

National Archives

" data-portal-copyright="

National Archives

" />

It’s not 1979 anymore. And now AI is changing the way people conduct research on humans, but our ethical and regulatory frameworks have not kept up.

Tamiko Eto, a certified IRB professional (CIP) and expert in the field of HSR protection and AI governance, is working to change that. Eto founded TechInHSR, a consultancy that supports IRBs reviewing research involving AI. I recently spoke with Eto about how AI has changed the game and the biggest benefits — and greatest risks — of using AI in HSR. Our conversation below has been lightly edited for length and clarity.

You have over two decades of experience in human subjects research protection. How has the widespread adoption of AI changed the field?

AI has actually flipped the old research model on its head entirely. We used to study individual people to learn something about the general population. But now AI is pulling huge patterns from population-level data and using that to make decisions about an individual. That shift is exposing the gaps that we have in our IRB world, because what drives a lot of what we do is called the Belmont Report.

That was written almost half a century ago, and that was not really thinking about what I would term “human data subjects.” It was thinking about actual physical beings and not necessarily their data. AI is more about human data subjects; it’s their information that’s getting pulled into these AI systems, often without their knowledge. And so now what we have is this world where massive amounts of personal data are collected and reused over and over by multiple companies, often without consent and almost always without proper oversight.

Could you give me an example of human subjects research that heavily involves AI?

In areas like social-behavioral-education research, we’re going to see things where people are training on student-level data to identify ways to improve or enhance teaching or learning.

In health care, we use medical records to train models to identify possible ways that we can predict certain diseases or conditions. The way we understand identifiable data and re-identifiable data has also changed with AI.

So right now, people can use that data without any oversight, claiming it’s de-identified because of our old, outdated definitions of identifiability.

Where are those definitions from?

Health care definitions are based on HIPAA.

The law wasn’t shaped around the way that we look at data now, especially in the world of AI. Essentially it’s saying that if you remove certain parts of that data, then that individual might not reasonably be re-identified — which we know now is not true.

What’s something that AI can improve in the research process — most people aren’t necessarily familiar with why IRB protections exist. What’s the argument for using AI?

So AI does have real potential in improving health care, patient care and research in general — if we build it responsibly. We do know that when built responsibly, these well-designed tools can actually help catch problems earlier, like detecting sepsis or spotting signs of certain cancers with imaging and diagnostics because we’re able to compare that outcome to what expert clinicians would do.

Though I’m seeing in my field that not a lot of these tools are designed well and nor is the plan for their continued use really thought through. And that does cause harm.

I’ve been focusing on how we leverage AI to improve our operations: AI is helping us handle large amounts of data and reduce repetitive tasks that make us less productive and less efficient. So it does have some capabilities to help us in our workflows so long as we use it responsibly.

It can speed up the actual process of research in terms of submitting an [IRB] application for us. IRB members can use it to review and analyze certain levels of risk and red flags and guide how we communicate with the research team. AI has shown to have a lot of potential but again it entirely depends on if we build it and use it responsibly.

What do you see as the greatest near-term risks posed by using AI in human subjects research?

The immediate risks are things that we know already: Like these black box decisions where we don’t actually know how the AI is making these conclusions, so that is going to make it very difficult for us to make informed decisions on how it’s used.

Even if AI improved in terms of being able to understand it a little bit more, the issue that we’re facing now is the ethical process of collecting that data in the first place. Did we have authorization? Do we have permission? Is it rightfully ours to take and even commodify?

So I think that leads into the other risk, which is privacy. Other countries may be a little bit better at it than we are, but here in the US, we don’t have a lot of privacy rights or self data ownership. We’re not able to say if our data gets collected, how it gets collected, and how it’s going to be used and then who it’s going to be shared with — that essentially is not a right that US citizens have right now.

Everything is identifiable, so that increases the risk that it poses to the people whose data we use, making it essentially not safe. There’s studies out there that say that we can reidentify somebody just by their MRI scan even though we don’t have a face, we don’t have names, we don’t have anything else, but we can reidentify them through certain patterns. We can identify people through their step counts on their Fitbits or Apple Watches depending on their locations.

I think maybe the biggest thing that’s coming up these days is what’s called a digital twin. It’s basically a detailed digital version of you built from your data. So that could be a lot of information that’s grabbed about you from different sources like your medical records and biometric data that may be out there. Social media, movement patterns if they’re capturing it from your Apple Watch, online behavior from your chats, LinkedIn, voice samples, writing styles. The AI system then gathers all your behavioral data and then creates a model that is duplicative of you so that it can do some really good things. It can predict what you’ll do in terms of responding to medications.

But it can also do some bad things. It can mimic your voice or it can do things without your permission. There is this digital twin out there that you did not authorize to have created. It’s technically you, but you have no right to your digital twin. That’s something that’s not been addressed in the privacy world as well as it should be, because it’s going under the guise of “if we’re using it to help improve health, then it’s justified use.”

What about some of the long-term risks?

We don’t really have a lot we can do now. IRBs are technically prohibited from considering long-term impact or societal risks. We’re only thinking about that individual and the impact on that individual. But in the world of AI, the harms that matter the most are going to be discrimination, inequity, the misuse of data, and all of that stuff that happens at a societal scale.

“If I was a clinician and I knew that I was liable for any of the mistakes that were made by the AI, I wouldn’t embrace it because I wouldn’t want to be liable if it made that mistake.”

Then I think the other risk we were talking about is the quality of the data. The IRB has to follow this principle of justice, which means that the research benefits and harm should be equally distributed across the population. But what’s happening is that these usually marginalized groups end up having their data used to train these tools, usually without consent, and then they disproportionately suffer when the tools are inaccurate and biased against them.

So they’re not getting any of the benefits of the tools that get refined and actually put out there, but they’re responsible for the costs of it all.

Could someone who was a bad actor take this data and use it to potentially target people?

Absolutely. We don’t have adequate privacy laws, so it’s largely unregulated and it gets shared with people who can be bad actors or even sell it to bad actors, and that could harm people.

How can IRB professionals become more AI literate?

One thing that we have to realize is that AI literacy is not just about understanding technology. I don’t think just understanding how it works is going to make us literate so much as knowing what questions we need to ask.

I have some work out there as well with this three-stage framework for IRB review of AI research that I created. It was to help IRBs better assess what risks happen at certain development time points and then understand that it’s cyclical and not linear. It’s a different way for IRBs to look at research phases and evaluate that. So building that kind of understanding, we can review cyclical projects so long as we slightly shift what we’re used to doing.

As AI hallucination rates decrease and privacy concerns are addressed, do you think more people will embrace AI in human subjects research?

There’s this concept of automation bias, where we have this tendency to just trust the output of a computer. It doesn’t have to be AI, but we tend to trust any computational tool and not really second guess it. And now with AI, because we have developed these relationships with these technologies, we still trust it.

And then also we’re fast-paced. We want to get through things quickly and we want to do something quickly, especially in the clinic. Clinicians don’t have a lot of time and so they’re not going to have time to double-check if the AI output was correct.

I think it’s the same for an IRB person. If I was pressured by my boss saying “you have to get X amount done every day,” and if AI makes that faster and my job’s on the line, then it’s more likely that I’m going to feel that pressure to just accept the output and not double-check it.

And ideally the rate of hallucinations is going to go down, right?

What do we mean when we say AI improves? In my mind, an AI model only becomes less biased or less hallucinatory when it gets more data from groups that it previously ignored or it wasn’t normally trained on. So we need to get more data to make it perform better.

So if companies are like, “Okay, let’s just get more data,” then that means that more than likely they’re going to get this data without consent. It’s just going to scrape it from places where people never expected — which they never agreed to.

I don’t think that that’s progress. I don’t think that’s saying the AI improved, it’s just further exploitation. Improvement requires this ethical data sourcing permission that has to benefit everybody and has limits on how our data is collected and used. I think that that’s going to come with laws, regulations and transparency but more than that, I think this is going to come from clinicians.

Companies who are creating these tools are lobbying so that if anything goes wrong, they’re not going to be accountable or liable. They’re going to put all of the liability onto the end user, meaning the clinician or the patient.

If I was a clinician and I knew that I was liable for any of the mistakes that were made by the AI, I wouldn’t embrace it because I wouldn’t want to be liable if it made that mistake. I would always be a little bit cautious about that.

Walk me through the worst-case scenario. How can we avoid that?

I think it all starts in the research phase. The worst case scenario for AI is that it shapes the decisions that are made about our personal lives: Our jobs, our health care, if we get a loan, if we get a house. Right now, everything has been built based on biased data and largely with no oversight.

The IRBs are there for primarily federally funded research. But because this AI research is done with unconsented human data, IRBs usually just give waivers or it doesn’t even go through an IRB. It’s going to slip past all these protections that we would normally have built in for human subjects.

At the same time, people are going to be trusting these systems so much they’re just going to stop questioning its output. We’re relying on tools that we don’t fully understand. We’re just further embedding these inequities into our everyday systems starting in that research phase. And people trust research for the most part. They’re not going to question the tools that come out of it and end up getting deployed into real-world environments. It’s just consistently feeding into continued inequity, injustice, and discrimination and that’s going to harm underrepresented populations and whoever’s data wasn’t the majority at the time of those developments.