80% of U.S. Doctors Use AI Tools Without FDA Approval
Updated
Updated · The Atlantic · Jun 17
80% of U.S. Doctors Use AI Tools Without FDA Approval
3 articles · Updated · The Atlantic · Jun 17
Summary
Eighty percent of doctors already use AI tools on the job, according to a 2026 American Medical Association survey, even as many generative products deployed in hospitals have not been cleared by the FDA for medical use.
That gap exists because many systems are marketed as clinical decision support or wellness tools, categories that can avoid FDA review if diagnosis and treatment formally remain with physicians.
Recent evidence cuts both ways: a Harvard- and Stanford-led study found ChatGPT beat physicians on diagnostic cases, but a randomized NEJM AI trial showed erroneous AI output could easily mislead doctors.
Hospitals and consumer apps are still pushing deeper into care, from drafting patient messages to interpreting lab results and linking medical records, while companies insist they provide information rather than medical advice.
Regulators and researchers are now debating whether medical chatbots should face traditional FDA authorization or even physician-like licensing, but critics warn adoption is moving faster than oversight.
Is your doctor's new AI assistant practicing medicine without a license?
As AI rewrites medicine, will patient safety be the price of rapid innovation?
AI can now diagnose patients and bill insurance, but who is liable when it's wrong?
AI in U.S. Healthcare: 1,016 Devices Approved Amid Regulatory Gaps, Safety Risks, and Calls for Reform
Overview
Artificial intelligence is rapidly transforming healthcare, with U.S. physicians increasingly adopting AI tools in their daily practice. This surge is driven by growing confidence in AI’s clinical benefits, as healthcare organizations use AI for diagnosing diseases, monitoring patients, and automating routine tasks. However, nearly three-quarters of physicians remain concerned about AI’s accuracy and reliability. Financial incentives from AI device vendors, especially in fields like radiology and cardiology, have also increased. Despite widespread use, many AI tools lack rigorous regulatory review, highlighting a significant gap between rapid adoption and the assurance of safety and effectiveness.