The Food and Drug Administration (FDA) holds a dual role with respect to artificial intelligence: it regulates AI/ML-enabled medical devices, and it deploys AI internally for agency operations including drug and device review. As a regulator, the FDA's approval pipeline determines which clinical AI reaches patients, connecting it to Healthcare — AI Deployment.
As regulator
As of 2025, the FDA had cleared more than 500 AI/ML-enabled Software as a Medical Device (SaMD) products. Since 2021 the agency has maintained an Action Plan for AI/ML-based SaMD, a framework intended to address continuous-learning AI in clinical settings.
A senior FDA official signaled, in remarks reported July 2, 2026, that the agency will soon seek public input on AI that can practice medicine, after clearing one such product through the 510(k) pathway and engaging with a company pursuing the more stringent de novo pathway (Source: insideaipolicy.com).
On August 19, 2026 the Digital Health Center of Excellence within the FDA's Center for Devices and Radiological Health published "Considerations for the Regulation of Generative AI-Enabled Medical Devices: Discussion Paper and Request for Feedback," seeking comment under docket FDA-2026-N-7874 by October 19, 2026 on risk assessment, premarket evaluation, postmarket monitoring and other topics, across 26 numbered questions (Considerations for the Regulation of Generative AI-Enabled Medical Devices (FDA CDRH, 2026)). The paper proposes a two-axis risk heuristic — device activity against the consequence of relying on an incorrect output — and a competency-based premarket approach, inspired at a high level by how clinicians are credentialed, combining non-clinical device benchmarking with clinical confirmation. Shadow deployment, in which a device runs in a live workflow without its outputs reaching clinicians or patients, is one of five listed confirmation approaches. It also raises voluntary Foundation Model Device Master Files, under which model developers could submit structured model or system cards to FDA confidentially, and puts CDRH on record about agentic AI systems in a premarket-evaluation context. The paper disclaims being draft or final guidance, states it does not communicate CDRH's proposed or final regulatory expectations, and does not address whether the approaches lie within FDA's existing legal authorities. The agency's move was first reported on August 18, 2026 as a "request for information on the regulation of generative artificial intelligence"; both the instrument class and the subject are wider in that account than in the document, which is a discussion paper on generative AI-enabled medical devices (Source: insideaipolicy.com). See Agentic AI.
The agency's draft agreement on Medical Device User Fee Amendments (MDUFA) reauthorization, cleared by the Office of Management and Budget on July 8, 2026, includes most digital-health priorities stakeholders sought, including health-technology sandboxes and a streamlined pre-submission process (Source: insideaipolicy.com).
AI uses and deployments
The FDA began publicly documenting internal AI deployments in 2025. On May 8, 2025, it announced completion of its first AI-assisted scientific review pilot, alongside an agency-wide AI rollout timeline ("FDA Announces Completion of First AI-Assisted Scientific Review Pilot and Aggressive Agency-Wide AI Rollout Timeline," FDA press release), marking the first time the agency used AI to assist in actual drug and device regulatory review. On June 2, 2025, the FDA launched an agency-wide internal productivity tool ("FDA Launches Agency-Wide AI Tool to Optimize Performance for the American People," FDA press release).
On April 29, 2026, the FDA announced a pilot it described as first-of-its-kind, using cloud computing and AI for real-time clinical drug trials. FDA Chief Artificial Intelligence Officer Jeremy Walsh said the pilot could lead to reductions of 20–40% in overall clinical trial duration (Source: nextgov.com). The same day, the Mayo Clinic disclosed its REDMOD system, an AI tool that flagged subtle changes in routine CT scans 475 days before a pancreatic-cancer diagnosis (Source: bloomberg.com).
On June 4, 2026, the FDA extended the pilot's public comment period by a month, following industry requests for operational detail on how the real-time, AI-assisted trials process would run (Source: insideaipolicy.com).
Policy positions and oversight
On April 24, 2026, HHS Secretary Robert F. Kennedy Jr. argued that the FDA's expanding use of AI tools had accelerated drug reviews and could eventually replace human reviewers and enable personalized medicine, framing AI as something that could make the FDA "irrelevant" (Source: insideaipolicy.com).
Acting Commissioner Kyle Diamantas's first public outline of agency priorities, which became public on July 21, 2026 following the departure of former commissioner Marty Makary, emphasized AI and digital tools to compete with foreign nations, improve nutrition, and protect medical products from shortages (Source: insideaipolicy.com).
The agency created a Deputy Commissioner for Technology and Artificial Intelligence, announced September 8, 2026, to be filled by Jared Seehafer, a senior advisor to the agency for the past year who played a role in crafting the FDA's white paper on generative AI-driven device regulation. A medical-device industry veteran who founded an electronic quality management system startup, Seehafer has advocated a hands-off regulatory approach for less-risky clinical decision support (Source: insideaipolicy.com). The report returned only its lede, so the new post's relationship to the existing Chief Artificial Intelligence Officer role is not established here.
Two cross-cutting concerns appear in commentary on the FDA's AI adoption. Tyler Cowen has argued that improved drug discovery does not accelerate availability if FDA review timelines remain slow (Why I Think AI Take-Off Is Relatively Slow (Cowen)); the agency's own AI adoption may speed review, which raises a separate concern about AI-assisted approval errors in a life-critical domain. Separately, Lancet Endoscopist Deskilling Study (2025) reports that clinical AI may erode practitioner skill; the FDA does not currently regulate for deskilling as a risk category.
Relationships
- regulated-by: Congressional oversight; HHS.
- related: Healthcare — AI Deployment, Lancet Endoscopist Deskilling Study (2025), AMA Physician AI Sentiment Report (2026), California AB 3030 (Healthcare AI Disclosure), Why I Think AI Take-Off Is Relatively Slow (Cowen), Government AI Procurement, Federal AI Adoption — Patterns and Tensions.
- related: Considerations for the Regulation of Generative AI-Enabled Medical Devices (FDA CDRH, 2026) — the August 2026 discussion paper on generative AI-enabled devices.