Health and Human Services Secretary Robert F. Kennedy Jr. has become an unlikely cheerleader for artificial intelligence in medicine — and, as The Bulwark’s Jonathan Cohn reports, the chatbots he’s championing don’t think much of his science. Speaking at a MAHA Action conference in Washington in late September, Kennedy declared that “it would be malpractice for a doctor to make a diagnosis, or to make a prescription, without at least checking AI.”
Kennedy pitched AI as a way for ordinary people to challenge doctors and scientists, suggesting it might reveal that COVID vaccines don’t do what officials claimed. AI, he promised, would “free us from medical tyranny.”
Cohn agrees AI could transform medicine, particularly in diagnostic imaging and fast-moving fields like oncology. “The data are pretty good that AI is more accurate than physicians at creating a differential diagnosis,” said Ezekiel Emanuel, a University of Pennsylvania oncologist and bioethicist, who co-authored a recent article in the Journal of the American Medical Association arguing that AI alone often outperforms doctors who merely consult it.
But AI’s knack for digesting scientific data also makes it a capable fact-checker. Cohn asked Anthropic’s Claude to evaluate four of Kennedy’s signature claims. On his assertion that only the COVID shot among 18 childhood vaccines received a placebo-controlled trial, and that “there was no safety study,” Claude found Kennedy inflated dose counts and misstated how vaccines are tested. “A trial without a placebo is still a safety study,” it noted. New versions of old vaccines are tested against existing ones because leaving children unprotected against diseases like measles would be unethical.
Related on the SLP
Claude found no credible evidence for Kennedy’s claim that a recent meta-review supported splitting up the MMR shot, concluding that research “points the other way.” It rated “false” his claim that SV40-contaminated polio vaccines triggered an “explosion of soft tissue cancers,” citing population data — including a study of 1.8 million Danish vaccine recipients — showing no excess cancer. And it rejected his description of COVID vaccines as “the deadliest ever made,” noting that reports to the federal Vaccine Adverse Event Reporting System can’t establish that a vaccine caused harm.
ChatGPT, Google’s Gemini and even Elon Musk’s Grok reached essentially the same verdicts, rating all four claims false or misleading. “The great thing about using AI to check medical information,” Emanuel told Cohn, “is that it shows RFK Jr. doesn’t know what he is talking about.”
