An AI Detective at HHS: Robert F. Kennedy Jr. Hunts for Side Effects Nobody Else Has Seen

Imagine handing a powerful digital microscope to someone who is already convinced that every woe in the world is caused by one particular kind of germ. He will find them, even if it is just dust on the lens or a flaw in the glass. That is roughly what is happening now in the corridors of the US Department of Health and Human Services (HHS).

The agency, now inseparable from the controversial figure of Robert F. Kennedy Jr., has decided to arm itself with artificial intelligence to look for links between vaccination and various diseases.

Data analysis in itself is a worthy and even necessary idea. In medicine there is always room to search for extremely rare side effects that may have slipped past researchers during standard clinical trials. But the devil, as usual, is in the details, and in who exactly holds the remote control for this technology.

For decades, Kennedy built his public career on hard skepticism toward vaccines. Now he has a tool that can turn scattered and often unverified complaints into supposedly science-based hypotheses. The problem is that modern AI is neither an impartial judge nor a bearer of absolute truth.

It is a mirror of the data fed into it and the prompts dictated to it. Feed a neural network the huge VAERS database, which collects every complaint people report after vaccination, from a mild headache to random injuries, and insistently ask it to find patterns, and it will find them. Neural networks are remarkably good at hallucinating correlations, especially when asked to.

Scientists around the world fear the new tool will become a conveyor belt for headlines about proven harm, ignoring the basic medical principle that after does not mean because of.

This case raises a crucial ethical question for the whole AI industry. We are used to discussing chatbots' absurd mistakes, such as mixing up celebrities' birth dates or inventing books that do not exist. But when an algorithm starts generating medical hypotheses that can influence government policy and the health of millions, the stakes soar. Using machine learning to confirm one's own bias (so-called confirmation bias) is perhaps the most dangerous way to apply technology in government. Instead of using algorithms to objectively check drug safety, we risk getting a perfect tool for legitimizing conspiracy theories.

The situation is complicated by the fact that the general public tends to trust conclusions drawn by a computer more than the words of politicians. If HHS starts issuing reports generated by a neural network, it will be extremely hard for the scientific community to rebut them. Arguing with numbers and charts backed by a mysterious and powerful intelligence is psychologically harder than arguing with a public figure. We are entering an era of alternative data, in which AI may become not a doctor's assistant but the main weapon in an information war against evidence-based medicine. And we will be dealing with the consequences of this experiment for a very long time.

The key question: Can the scientific community build a system of checks and balances, or are we heading into an era in which medical standards are dictated by algorithms with a predetermined search direction?