RFK Jr. says AI can challenge public-health advice

At the September 29 MAHA Summit, U.S. Health Secretary Robert F. Kennedy Jr. said artificial intelligence could give patients medical second opinions and help them challenge guidance from doctors and public-health officials. The remarks put general-purpose AI in a role that could affect patient decisions, although the opened record identifies no system, clinical-use boundary or safety plan.

Kennedy also said the Department of Health and Human Services was working to make AI easy for Americans to use and that every American would have medical records available on a cellphone by the end of the administration. The event transcript provides no implementation plan, security design or details about which AI system would be used.

Kennedy said an AI system could review and distill a person’s medical history: “It can give you a second opinion that is much better informed than any doctor in the country.” He offered no evidence for that comparison. The opened record does not specify the model, its intended users, its accuracy threshold, a process for clinician oversight or a way to report harmful outputs.

He also said people could use AI to check advice from public officials, invoking guidance about masks, social distancing and vaccines before saying the technology could “free us from medical tyranny.” Those remarks framed a general-purpose AI system as a counterweight to medical and public-health expertise, but established evidence does not support a categorical dismissal of those measures.

A systematic review and meta-analysis published in BMJ associated mask wearing with lower COVID-19 incidence, with a pooled relative risk of 0.47 and a 95% confidence interval of 0.29 to 0.75. It associated physical distancing with lower incidence as well, though the review reported substantial differences among studies. Current CDC guidance says masks can reduce transmission of respiratory viruses and that protection depends on mask type and fit.

A study of vaccinated people and their contacts found that two vaccine doses were associated with reduced onward transmission of the Alpha and Delta coronavirus variants in England. The reduction was smaller for Delta and waned over time, so the findings do not establish the same effect for every vaccine, variant or period.

Using general-purpose AI for individual health guidance introduces a separate set of safety and policy questions. The World Health Organization’s guidance says generative systems used for diagnosis, clinical care or patient questions can produce false, inaccurate, biased or incomplete statements. It also warns of automation bias, in which patients or clinicians give too much weight to a system’s output, and of cybersecurity risks to patient information and the integrity of health algorithms.

WHO recommends that governments set standards for health-related large multimodal models, assign regulators to assess and approve them where resources permit, and require independent audits and impact assessments after release for large-scale deployments. Those recommendations make the proposed combination of AI and cellphone access to medical records a matter of data governance as well as clinical safety.

The regulatory treatment would depend on what a system is designed to do. FDA guidance issued in January says only certain clinical decision-support functions fall outside the statutory definition of a medical device. Existing digital-health policies continue to apply to software that meets that definition, including software intended for patients or caregivers.

HHS has not identified in the opened record whether Kennedy’s proposal is a formal program, which product it would use, how patient consent and access would work or what clinical and privacy safeguards would apply.