If the FDA isn’t regulating every aspect of AI in healthcare who will? Executives debate at San Diego waterfront – San Diego Union-Tribune

Some of the biggest names in biotech, pharma, venture capital and AI came to the San Diego Convention Center this week — but the most important conversations happened behind closed doors.

At the Marriott Marquis, about a block away from the packed convention floor, 88 CEOs and senior executives spent hours trying to answer one question: What does AI mean for the future of global healthcare?

“It’s not going to just take one person, one group, or one entity. It has to be all of us moving together,” said Capt. Dianne Paraoan, acting director of the Office of Medical Policy at the Food and Drug Administration. She flew in from Washington to attend BIO International, the world’s largest gathering of the biotechnology and life sciences industry.

But the tech industry didn’t wait for an answer. It just showed up with an AI scientist.

“All of a sudden you’ve got this massive disruptive technology,” said John Crowley, CEO of the Biotechnology Innovation Organization, the industry’s largest trade group. “It is going to fundamentally change business processes, the way in which we think, the way in which we work — but it’s also a challenge, because in our industry we work on such long timelines.”

During Nvidia’s private panel, the executives listed dozens of companies that they’ve worked with to deploy AI agents already testing molecules and running experiments — and some are outperforming human scientists in the lab.

Nvidia’s agentic model can create new molecules, predict whether a drug will do well in clinical trials, and run simulated experiments in a fraction of the time actual scientists can.

Though it’s worth noting that these models, at least now, only save time, not money — compute is becoming increasingly expensive.

“When you let them (AI scientists) run for days, you can actually accomplish months of intellectual work,” said Andrew White, chief technology officer and co-founder of San Francisco-based Edison Scientific, a firm that created its own AI scientist named Cosmo.

White has yet to deploy — or employ, if you believe in anthropomorphizing AI — Cosmo in an actual lab.

But San Francisco-based Chai Discovery is working to ship its agentic models now, inking over $1 billion in big pharma deals this year alone. Chai built a model that can design molecules and antibodies for drug discovery.

A year ago, finding a viable antibody meant screening a thousand molecules to get one that worked. “Today that success rate has jumped to 10% to 15%, two orders of magnitude in 12 months,” said Josh Meier, CEO and co-founder of Chai.

And it’s this pace that the FDA is excited about.

On Monday, the Department of Health and Human Services and the Food and Drug Administration announced a broader program called “Operation Trailblazer,” a series of measures designed to speed up drug development from the early investigational stage through late-stage clinical trials.

The FDA seems to be loosening guidelines while creating new AI frameworks. For example, earlier this year, the agency decided that recommendation-based AI tools did not require premarket review when being used with clinician oversight.

At the same time, the FDA published a seven-step risk-based system for determining whether an AI model is reliable enough to inform high-stakes regulatory decisions — such as whether a drug is effective or manufactured properly, Paraoan explained.

But while this framework is “useful” according to those building AI software — like Samuel Aronson, chief AI officer at Boston-based healthtech firm AIwithCare — much of the oversight is done by the companies themselves.

“FDA isn’t going to regulate what we haven’t regulated,” Paraoan said. Because what the AI scientists are doing in the lab doesn’t directly impact patients, the FDA has no technical jurisdiction over the lab work being done. Yet, as these models increasingly enter clinical trials, the agency is under pressure to create more concrete guidelines.

“That’s why I’m here today,” said Paraoan. “To have those discussions.”

Almost every AI healthcare company at the conference maintained that safety was their “No. 1 priority,” including Tran Le, general manager of Palo Alto-based Hippocratic AI, a company that uses AI agents to educate and enroll prospective patients in clinical trials.

“I think that’s a good thing,” Paraoan said. “Because again, we’re not going to regulate something we don’t normally regulate. If it’s related to patient safety, we care about it … but the guardrails are something these companies need to think about to make sure that it’s not just FDA, but the public is comfortable with the tools.”

After the conference, a few members of the audience questioned whether the self-given safety benchmarks were legitimate. “It’s easy to pass a test that you write,” one attendee said.

During the Nvidia panel, Powell told the Union-Tribune that the new agentic life science toolkit is available on GitHub for anyone with ChatGPT or Claude to begin experimenting with, but “if it becomes a concern, we will, of course, abide by whatever the concern is, but at this point, being able to sequence a genome and get a clinical variant in the matter of minutes versus days is very good for humanity.”