AI chatbots outperform doctors on medical image diagnosis at a statistically significant level. New York's response? Make it illegal for them to answer health questions.
Senate Bill S7263, sponsored by Sen. Kristen Gonzalez (D-Queens), reached the New York Senate floor calendar on February 26, 2026. The bill would hold AI companies liable whenever their chatbots provide "substantive responses" in any of 13 licensed professions: medicine, law, dentistry, nursing, psychology, social work, engineering, architecture, pharmacy, optometry, podiatry, physical therapy, and veterinary medicine.
That covers OpenAI, Anthropic, Google, xAI, and essentially every company running a general-purpose chatbot accessible in New York.
The Numbers That Make This Awkward
The timing is hard to ignore. A study using the New England Journal of Medicine's Image Challenge found that all tested language models "significantly outperformed" physicians at p < 0.001. In neuroradiology specifically, Claude 3.5 scored 80.4% accuracy compared to 71.4% for first-year fellows and 51.8% for junior faculty. A meta-analysis covering 83 studies found "no statistically significant difference" between AI and non-expert physicians in diagnostic accuracy.
Meanwhile, the American Medical Association reported in 2026 that physician AI usage has doubled from 38% in 2023 to 81% today. Doctors are already using these tools. The bill would restrict what patients can access directly while physicians quietly rely on the same technology behind closed doors.
Who Wants This and Why
The bill has backing from the AFL-CIO and the Working Families Party. Mario Cilento of the AFL-CIO supports it as worker protection against AI displacement. That framing makes the politics clear: this is partly a labor play dressed up as consumer safety.
Critics see something more blunt. Taylor Barkley of the Abundance Institute calls it protectionism for incumbents' billable hours. The argument has teeth when you look at the access problem. New York has over 900,000 uninsured residents. An estimated 92% of low-income legal issues go underserved. Healthcare costs have risen from $2,900 per person in 1990 to $14,570 in 2023. Medical errors remain the third leading cause of death in the U.S., killing more than 250,000 people annually.
Banning AI from answering medical or legal questions doesn't fix any of those problems. It arguably makes them worse by removing the most accessible option for people who can't afford a specialist.
The Regulatory Pattern
New York already enacted the RAISE Act in early 2026, and S7263 has a companion bill in the Assembly (A6545). This isn't a one-off proposal. It's part of a growing state-level push to regulate AI through professional licensing frameworks rather than safety standards.
California tried something similar with SB 1047, which Governor Newsom vetoed. The FDA already has a medical device approval process that could cover AI diagnostic tools. The question is whether states will build on existing regulatory infrastructure or create a patchwork of profession-by-profession bans.
The practical effect of S7263 would be strange. ChatGPT could help you write a novel but couldn't suggest you see a doctor about chest pain. Claude could debug your code but couldn't explain what a prescription drug interaction means. The bill doesn't distinguish between "AI pretending to be a doctor" and "AI sharing publicly available medical information."
None of this means AI medical advice is risk-free. But the bill's approach of blanket liability across 13 professions treats a nuanced problem with a sledgehammer. The people most likely to lose access are the ones who couldn't afford a professional visit in the first place.