Former FDA Commissioner Robert Califf on RFK Jr., the FDA Exodus, and China

Robert Califf is the only person to have served twice as Commissioner of the U.S. Food and Drug Administration, first in 2016 and again from 2022 to 2025. Vid Desai spent six years modernizing the agency’s technology, first as Chief Technology Officer and then as the first Chief Information Officer in FDA history to report directly to the Commissioner. Both left in 2025. On this episode of Open Door Salon, they describe what they see happening to the agency now, and what it is costing the people who develop medical products.

What the turnover actually costs

Host Lori Ellis opens with the churn as she counts it: “six different CBER directors, six different CDER directors, we have an acting commissioner right now.” Her question is what that kind of turnover does to an agency, not only in the moment but over the years that follow.

Califf does not soften his answer. He calls it “calamitous for the United States and for the FDA,” then reframes the disagreement as a question about the civil service itself. “If you believe that there is value in having a civil service that provides continuity and standards and methods of doing things that are independent of politics, then I think you’d have to believe that we’re undergoing a very dangerous time right now.” He points to Project 2025 as the place where the opposing view is written down plainly, and notes that the document is clear about tearing the agencies down and silent on how they get rebuilt.

Desai’s concern is operational. The institutional knowledge, he says, “is obviously pretty much gone,” and it is not a line item that can be refilled by hiring. He arrived at the FDA from private industry and took six months to find his footing, a period he describes as genuinely difficult. Multiply that by thousands of positions turning over with every change of administration, he argues, and the agency never reaches the point where it can do anything significant.

Califf sharpens what this means for industry specifically. Most of the consequential work at the FDA happens long before a marketing application is filed, in the meetings where sponsors and reviewers work out what evidence will count. “If the people at FDA have no experience and they don’t understand how the system works, how can you bank on what they say five years later when you have to bring all this to the forefront to put your application in?”

Referees, coaches, and political overruling

Asked whether he expects scientific decisions to be overruled for political reasons, Califf says he has no inside knowledge of the current administration but describes what it looks like from outside as a different philosophy entirely, a view that all these decisions should be political. He goes further: “It looks somewhat like a kleptocracy, that is the political oversight is heavily influenced by who’s giving money and who’s benefiting, as opposed to the scientific merit.”

He offers a framing that runs through the rest of the conversation. The FDA is a referee. It does not own the team and it does not write the rules of the game, which are set by laws passed by Congress. Partway through he corrects his own analogy, and the correction is the sharper version: “I don’t see the commissioner as a referee. I see the civil servants as referees. I see the commissioner is more like the coach.” Referees who do not know the rulebook make bad calls, and it is the coach’s job to argue them.

On the national priority voucher program, Califf’s objection is procedural rather than ideological. He supports incentives for genuine national priorities, but he wants the priority list set by rule and applied by impartial civil servants, not decided case by case by political appointees. “If we prioritize everything then nothing has priority.” He is blunt about the failure mode: a celebrity making a phone call that turns into a priority designation is not a system.

He also flags a quieter problem, one he thinks people interested in FDA history should read for themselves. A recent reversal memo justified overturning a predecessor’s decision on the grounds that “my predecessor had overruled the civil servants twice but there is no record of why she did it.” The absence of an administrative record, Califf notes, left nothing to argue about.

Elsa, AI, and the money underneath it

Ellis raises Elsa, the FDA’s generative AI tool, which was announced and then found within months to be hallucinating. Desai’s answer complicates the easy story. The Elsa groundwork started while he and Califf were still at the agency, including model testing and proof-of-concept work with the Oncology Center of Excellence. The problem he sees is not the technology but the funding structure beneath it.

Temporary money, he explains, buys impressive things and then runs out. “AI is great, but it’s not cheap… eventually those funds go out and then what do you do?” Roughly 40 to 50 percent of an initial investment is needed just for day-to-day operations, and without a plan to replace temporary funds the program shuts down. He has watched it happen to COVID programs and infant formula oversight programs already.

Califf pushes back on the premise that AI agents reduce headcount. He expects the opposite at the FDA, because the agency has been underfunded for so long that freeing staff from routine work surfaces demand that was always there and never met. He points to the last two user fee negotiations, where the biggest ask from mature drug and device companies was not faster timelines but more meetings, and the agency simply did not have the people to hold them.

Desai closes the section with an example from outside healthcare. IKEA found its AI could answer about half of routine call center questions, then looked at the half it could not answer and discovered most were requests for design help. Rather than cut staff, the company retrained them into a new design service. His conclusion is about the question, not the tool: ask an AI how to cut costs and it will tell you, ask how to grow revenue and you get a different answer entirely.

China, Operation TrialBlazer, and the missing strategy

The last policy section is the one with the sharpest numbers. Desai says that last year, for the first time, China’s research funding exceeded that of the United States, roughly $1.3 trillion against about $1.2 trillion, at the same moment U.S. federal research funding is being cut. “We’re entering a fight with literally our hands tied behind our back,” he says.

Califf’s read is built on direct experience. He helped start a joint American-Chinese university outside Shanghai and ran clinical trials in the country. What strikes him is the absence of internal conflict of interest: a unitary national mission to lead in biotechnology, against a fragmented and litigation-wary American system. “You can do a phase one trial in China for a third the cost and 6 months quicker than you can in the US and Europe.”

On Operation TrialBlazer, the HHS initiative to accelerate U.S. clinical research, Califf is supportive of the attention and skeptical of the target. The bottleneck, he argues, is not the FDA. It is research sites that are slow, expensive, and bureaucratic, structured to protect revenue and guard against lawsuits.

His conclusion is the line that should worry anyone building a China strategy around containment: “trying to wall ourselves off from China is a doomed strategy. It won’t work… we got to figure out how to coexist with China. We can’t, we don’t have the power to hold China back.” He adds, plainly, that he wants to be friends and not enemies.

Why they did it

Ellis closes with two questions she asks every guest. Califf traces his career to a specific moment: taking over an intensive care unit the same year the first angiogram was performed on a patient having a heart attack, when you could see the clot and had nothing to treat it with. “We basically had prayer, morphine, and nitrates. That was pretty much it for treatment.” The hardest decision he names is a Christmas Eve when four consecutive patients died in front of him in the middle of an experimental protocol, and he had to decide to keep going.

Desai’s hardest decisions are quieter: the constant tension between announcing work early and waiting until it is ready. He calls his six years at the FDA the best of his career and the lowest paid, and says the one job he would consider returning to is the one he left unfinished.