FDA cleared 295 AI devices this year
- The U.S. Food and Drug Administration’s running list of AI-enabled medical devices has surged past 1,200 products, after the agency said in January 2025 it had already authorized more than 1,000. - The FDA’s latest policy shift is procedural, not a new mass approval: final guidance now lets some AI device makers pre-negotiate future software updates through Predetermined Change Control Plans. - The count keeps rising while the rules are still being written, with separate draft guidance for lifecycle oversight and bias, transparency, and postmarket monitoring. (fda.gov)
The Food and Drug Administration has not announced a one-year clearance of 295 AI devices, but its official list now spans more than 1,200 AI-enabled medical devices. (fda.gov 1) (fda.gov 2) That list covers devices the agency says are authorized for marketing in the United States through standard device pathways, including 510(k) clearance, De Novo authorization, and premarket approval. (fda.gov) The FDA said on January 6, 2025 that it had authorized more than 1,000 AI-enabled devices. By November 6, 2025, a Center for Devices and Radiological Health official said the total had climbed to more than 1,200. (fda.gov 1) (fda.gov 2) Most of these products are not chatbots talking directly to patients. They are software tools that help read scans, flag urgent findings, guide imaging, or support clinical workflow. (fda.gov 1) (fda.gov 2) The regulatory issue now is less whether the FDA reviews AI devices at all than how it reviews software that changes after launch. Ordinary medical devices are relatively fixed; AI systems can be retrained, tuned, and updated. (fda.gov) (fda.gov) The agency’s answer is the Predetermined Change Control Plan, or PCCP. In final guidance posted in December 2024 and issued in August 2025, the FDA said manufacturers can spell out planned modifications, validation methods, and impact assessments in advance. (fda.gov) (fda.gov) If the FDA accepts that plan as part of a marketing submission, some later software changes can be made without a fresh submission for each update. The guidance applies across the 510(k), De Novo, and premarket approval pathways. (fda.gov) At the same time, the FDA is still writing broader rules for how developers should document safety, effectiveness, bias controls, transparency, and postmarket monitoring over an AI product’s full life cycle. That framework remains draft guidance as of April 26, 2026. (fda.gov) (fda.gov) The official device list itself comes with a caveat. The FDA says it is not comprehensive and is built mainly from AI-related terms in public authorization summaries and device classifications. (fda.gov) That means the headline number is best read as a floor for visible, FDA-identified AI devices, not a complete census of every clinical software product using machine learning. The faster that list grows, the more the debate shifts from whether AI is entering medicine to how closely each tool is tested after it arrives. (fda.gov) (fda.gov)
