RFK Jr Wants US Health Data Pooled for AI as Screen Time and Uber Studies Land

The pitch was broad. Kennedy runs the US Department of Health and Human Services and has spent years questioning vaccine safety. He said records from doctor visits, exercise and insurance claims should follow patients and be open to government and outside researchers. "We need to use America's health data to determine whether the food we eat, the chemicals we encounter, the vaccines that we're administering, the other environmental and lifestyle exposures are contributing in one way or another to the chronic disease epidemic," he said, according to the Guardian's write-up on 30 September.

Kennedy singled out Medicaid as the most useful pool, describing "hundreds of millions of lives" and "tens of millions of children who've been in that program for 10 or 20 years." Medicare, he said, supplies data on older Americans and people with disabilities. He also named the commercial platform HealthVerity, which links claims, prescriptions, lab results and electronic health records.

State reporting to the CDC, by contrast, was "unusable" in his telling, because states collect different fields. The Guardian reported that about two dozen people watched the livestream. The Maha Institute is headed by Mark Gorton, who described himself two weeks earlier as "the boldest anti-vaxxer in the anti-vax conference." Tony Lyons co-founded the institute, and his publishing house gave $4m to Kennedy through Maha foundations, according to New York Times reporting cited by the Guardian.

Two things sit oddly next to each other here. The first is the scale of what is being proposed: a national pool of health records, queried by AI, at an agency whose secretary has spent decades disputing the safety record of vaccines. The second is the state of the evidence he wants that machinery to settle.

On the same day, CNN published a review of decades of screen-time research covering nearly three million children aged 18 and under across 64 countries. The senior author, Sheri Madigan of the University of Calgary, told CNN the link to poorer outcomes in thinking, learning, speech, behaviour and emotional regulation was real but modest. "Most of the associations were relatively small, like most other predictors of child development, but small does not mean meaningless," she said. She also warned the findings are correlational: struggling children may turn to screens, rather than screens causing the struggles.

Two outside academics quoted by CNN backed the method and trimmed the claims. Tatiana Iniguez Berrozpe of the University of Zaragoza called it "a methodologically very sound article" drawing on an "extremely large" sample, while saying the association is very small. Esther van Sluijs of Loughborough University said it was high quality but could not draw conclusions about specific uses such as social media or gaming, because there was not enough data.

That is the pattern across this batch of research: big datasets, careful authors, hedged conclusions. Which is roughly the opposite of how the data debate in Washington is being framed.

Plastic chemicals, and a rider study with a rebuttal

Also on 30 September, the Environmental Justice Foundation published a summary of its Poison Plastics report, built on new analyses of the PlastChem dataset. The foundation argues that plastic-associated chemicals are present in most everyday items, from food packaging and takeaway containers to clothing, furniture, hygiene products, infant feeding bottles and children's toys, and that the case for safe-by-design products has not been met. The summary is a report abstract rather than a peer-reviewed study, and it does not put a number on human exposure in the text provided.

Business Insider took a different angle on the same day, publishing an analysis by Len Sherman, an executive in residence at Columbia Business School. Sherman examined 37,500 trips by drivers in six US cities, using data from GigU, an app that shows drivers per-minute and per-mile pay before they accept a trip. He found the time between matching and pickup rose 19% between the first quarter of 2023 and the same period in 2026, while the price per mile Uber charged rose 53% on average. Wait times rose in five of six cities, including Atlanta, Dallas and Houston, and fell slightly in Tampa.

Uber pushed back. "This analysis relies on a number of inaccuracies, which we've publicly refuted," a spokesperson told Business Insider, declining to address the wait-time and per-mile figures directly and pointing to a January blog post disputing that profitability came from raising prices while taking a larger share of fares. Sherman conceded a methodological limit: because he used driver-side data, the analysis excludes the time riders spend waiting to be matched with a driver, which he called "by definition, an underestimate."

The three stories do not share a subject, but they share a shape. Each puts a number on something people already suspect, each carries a caveat, and each arrives as the policy conversation around health data is getting louder.

Kennedy's own framing leans hard on speed. "Studies that used to take years, you can do them literally in seconds," he said, describing AI that can find patterns across millions of histories. He said OpenAI co-founder Sam Altman met him that week and told him HHS uses generative AI more than any other federal agency, a term Kennedy rendered as "superintelligence." He listed work already under way: an autism registry, studies on diet, electromagnetic fields, screen time and vaccines, and use of the Vaccine Safety Datalink to compare vaccinated and unvaccinated outcomes, examine aluminium adjuvants, hepatitis B timing in infancy, and influenza, Covid and HPV vaccines in children.

What the dossier does not contain is any published result from those efforts. The screen-time review, the plastics summary and the Uber analysis all come with authors willing to name their limits. The health data programme, as described, does not yet come with a study at all.