RFK Jr. says AI gives 'better informed' second opinions than doctors

Why it matters: Kennedy didn't just praise AI in general terms. He cited a specific claim from OpenAI CEO Sam Altman, saying Altman told him it would now be "malpractice" for a doctor to make a diagnosis without checking it against AI first.

Kennedy went further, framing this as part of a broader shift away from deferring to institutional authority. He said health policymaking will no longer be dominated by officials who simply tell the public to "trust the experts."

• Event: the Make America Healthy Again summit, Waldorf Astoria, September 29, 2026.
• Organizer: the MAHA Center, a tax-exempt organization.
• Other speakers: Vice President JD Vance, Agriculture Secretary Brooke Rollins, EPA Administrator Lee Zeldin, and CMS Administrator Mehmet Oz.

JD Vance echoed a similar theme in his own remarks, saying medical experts "are not the experts in the same way; they don't have the same control or monopoly on knowledge."

He shared an anecdote about attending an event where a man described using AI to help find a cancer treatment for his wife, after doctors had told the family nothing more could be done.

In real life stories like that one are genuinely powerful, and they're part of why AI in medicine has real, serious appeal. But one moving anecdote doesn't tell you how the technology performs across millions of different patients and conditions.

Background: Kennedy's comments about no longer deferring to officials who say "trust the experts" don't exist in a vacuum. Since being sworn in as HHS Secretary in February 2025, he has made a series of significant changes to federal vaccine policy and its oversight structure.

In June 2025, Kennedy removed all 17 members of the CDC's vaccine advisory committee and replaced them with his own picks. That reconstituted committee later voted to drop the longstanding recommendation that all newborns receive the hepatitis B vaccine, among other changes to the childhood immunization schedule.

What's next: those changes didn't go unchallenged. A federal judge ruled in March 2026 that Kennedy's overhaul of the advisory committee wasn't lawful, and specifically invalidated the panel's earlier votes on hepatitis B and COVID-19 vaccine recommendations. That legal and institutional back-and-forth is the direct context for Kennedy's framing of AI as an alternative source of medical authority, one he's positioning against the same public health establishment he's been actively restructuring.

The catch: research on AI's medical reliability paints a more mixed picture than the summit's framing suggested. AI performs strongly in specific, narrow areas, radiology, dermatology, and ophthalmology among them, sometimes matching or exceeding specialist accuracy on imaging tasks.

But that strength doesn't hold evenly across medicine. One study tested leading AI models with doctor-designed clinical scenarios that had deliberate factual errors planted inside them. The models repeated or built on those planted errors in up to 83% of cases, an error rate driven by flawed reasoning, not just missing information.

By the numbers: real-world harm has already been documented, not just as a theoretical risk. Reported cases include a person developing bromism, a form of chemical toxicity, after following health instructions generated by ChatGPT, and a transplant patient who ultimately lost their transplanted organ after stopping their prescribed medication based on misleading AI-generated advice they received.

Despite those documented risks, AI adoption among physicians has grown fast. About two-thirds of US doctors used AI tools for clinical tasks in 2024, up from 38% the year before, according to industry survey data.

Who's affected: that gap, rapid adoption alongside real documented failure cases, is the actual state of AI in medicine right now. It's neither the flawless authority Kennedy described nor a technology doctors are avoiding. It's a tool being adopted quickly, with real benefits in specific areas and real risks that haven't been fully solved.

Who's affected: Kennedy's account of Altman's "malpractice" comment is worth putting next to Altman's own broader public statements on the same topic. OpenAI has launched a feature called ChatGPT Health, letting users connect medical records and apps like Apple Health, which the company has described as meant to support general well-being rather than replace diagnosis and treatment.

Altman himself has said he believes ChatGPT is now better at diagnostics than most doctors worldwide. But he's also said, separately, that he personally would not trust ChatGPT with his own "medical fate" without a human doctor involved in the loop, calling that preference something that might make him "a dinosaur."

The catch: that's a more cautious position than Kennedy's summit remarks suggested. OpenAI is also currently facing a lawsuit filed in mid-2026 alleging ChatGPT gave a user "unauthorized" and harmful medical advice, a case that remains unresolved and doesn't itself prove or disprove any claim made at the summit, but underscores that the company building this technology hasn't settled the question of how much to trust it either.

Not entirely. Some MAHA movement supporters reportedly expressed frustration with the summit itself, feeling that its corporate sponsorships and heavy AI focus drifted from the movement's original priorities around pesticides and food chemical safety.

What's next: that internal friction is genuinely worth closely watching as it unfolds in the months ahead. A political movement built partly around distrust of institutional authority is now being offered a genuinely different kind of authority, AI systems, as a replacement, and not everyone within that movement seems fully convinced that's the right trade to make.

• AI can be a genuinely useful second opinion tool, particularly for imaging-based conditions where it performs strongly.
• It is not a reliable replacement for a doctor's judgment across all conditions, especially given documented cases of AI repeating factual errors in clinical scenarios.
• Real people have been harmed by following AI medical advice without a doctor's involvement. Treat AI health information as a starting point, not a final answer.
• This is now part of a live political conversation, not just a technology debate, so expect the rhetoric around AI and medical authority to keep getting louder.