In a move that signals the government’s growing appetite for artificial intelligence, the Food and Drug Administration has been holding closed-door discussions with OpenAI, the company best known for ChatGPT, about the future of drug evaluation in the United States.
The meetings—confirmed by multiple sources familiar with the talks—point to a broader push by the FDA to modernise and possibly accelerate its notoriously sluggish drug approval process.
For patients awaiting life-saving treatments, that could sound like hope. But for those watching the entanglement of big tech with public health, it also raises uncomfortable questions.
At the center of the conversation is a project known as cderGPT, a likely reference to the FDA’s Center for Drug Evaluation and Research (CDER).
According to insiders, the meetings have included a small team from OpenAI, along with individuals connected to Elon Musk’s somewhat opaque “Department of Government Efficiency.”
These players, often seen hovering at the nexus of Silicon Valley ambition and bureaucratic experimentation, have met several times in recent weeks with FDA officials.
The talks are being led by Jeremy Walsh, the FDA’s newly appointed—and first ever—AI officer. Walsh, sources say, has also been in dialogue with Peter Bowman-Davis, a 21-year-old Yale undergraduate currently on leave and acting as the chief AI officer at the Department of Health and Human Services.
Bowman-Davis’s unconventional appointment was first revealed by Politico, and he’s reportedly working under the influence of VC firm Andreessen Horowitz’s “American Dynamism” strategy—a program that explicitly aims to inject startup culture into government infrastructure.
Officially, no contracts have been signed. But behind the scenes, the outlines of a techno-regulatory revolution are beginning to take shape.
“Why Are We Not Modernised with AI?”
FDA commissioner Dr. Marty Makary, a vocal advocate for reform, took to X (formerly Twitter) this week to ask, “Why does it take over 10 years for a new drug to come to market?
Why are we not modernised with AI and other things?” He went on to reveal that the FDA had just completed its first-ever AI-assisted scientific review—though details remain scarce.
Makary stopped short of naming OpenAI or any specific models. Still, the timing of his comments, just days after his appearance at the American Hospital Association’s annual meeting, suggests the AI integration is more than just speculative.
He spoke enthusiastically at the event about AI’s potential in advancing treatments for diseases like diabetes and cancer.
Meanwhile, former commissioner Robert Califf—who led the agency during the Obama and Biden administrations—offered a more measured take. In an email response, he noted that FDA review teams have “been using AI for several years now.”
However, he did raise a pointed question: What exactly does ‘AI-assisted’ mean in this context? It’s a question no one at the agency has publicly answered.
A “Small Part” of a Much Bigger Machine
Even among AI advocates, there’s a consensus that using algorithms for final drug approval is only a sliver of what’s possible.
The majority of experimental drugs never make it to the FDA for review, falling victim to earlier-stage failures in trials or development. So while trimming the bureaucratic fat at the end stage could help, it’s hardly the silver bullet.
Rafael Rosengarten, CEO of Genialis and board member of the Alliance for AI in Healthcare, says he supports AI integration—in theory. But he warns that policy, transparency, and ethical training data must come first. “
Rosengarten says these machines are incredibly adept at learning information. “But they have to be trained in a way so they’re learning what we want them to learn.” he said.
The Risks of Speeding Up Science
If the public has learned anything from recent years, it’s that technology moves faster than regulation. That can be a blessing when innovation brings about better outcomes. But in the context of human health and drug safety, cutting corners for the sake of efficiency is a gamble no one can afford.
And here lies the tension. On one side: a bureaucratic agency that’s long been criticised for being slow, risk-averse, and analog. On the other: a private AI firm backed by Silicon Valley giants, known more for creating chatbot hallucinations than scientific accuracy.
The question is not whether AI will become a part of regulatory medicine—it already has. The question is who is steering the process, how decisions will be made, and whether the public will ever get a clear view of what’s under the hood.
For now, the FDA’s partnership with OpenAI is just a series of meetings, not a signed deal. But the seeds have been planted. And as the agency inches closer to automation, it owes the public something more than vague promises and cryptic tweets.
