The US Food and Drug Administration has now authorized more than 1,600 AI-enabled medical devices, according to the agency’s own public tracking list, a milestone that illustrates how thoroughly machine learning has woven itself into everyday clinical practice over the past decade. The pace of authorizations has only accelerated: an analysis of FDA data found 335 AI-enabled devices cleared in 2025 alone, with 181 more cleared in just the first six months of 2026.
What Counts as an AI Medical Device
The FDA’s authorized device list spans an enormous range of clinical applications, from AI software that flags suspicious regions on mammograms and CT scans to algorithms that monitor cardiac rhythms, predict sepsis risk, and assist radiologists in triaging which scans need urgent review. Most of these devices go through the FDA’s 510(k) clearance pathway, which requires manufacturers to demonstrate their product is substantially equivalent to an already-approved device, a lower bar than the full premarket approval process required for entirely novel medical technologies. The agency itself has cautioned that its public list, while extensive, is not a comprehensive inventory of every AI-enabled device on the market, meaning the true number in clinical use could be higher still.
Recent Examples Show the Breadth of Applications
Individual clearances illustrate how varied this category has become. DeepHealth received 510(k) clearance for an AI-powered breast ultrasound technology in August 2026, aimed at improving early cancer detection. Separately, a product called UpDoc reportedly became the first FDA-cleared software-as-a-medical-device to use patient-facing large language model technology, representing a notable shift from AI tools used behind the scenes by clinicians toward AI systems that interact directly with patients, a category regulators are still developing frameworks to evaluate.
Why Radiology and Imaging Dominate the List
A disproportionate share of FDA-authorized AI devices fall into radiology and medical imaging, reflecting both the maturity of image-based machine learning techniques and the relatively straightforward regulatory pathway for tools that assist, rather than replace, a radiologist’s final judgment. This pattern mirrors broader clinical validation trends: a Swedish trial found AI could flag breast cancer signs up to six years before diagnosis, while the UK’s GEMINI study showed AI boosted breast cancer detection rates by 10.4% and cut wait times to three days, both studies reinforcing imaging as the area where AI has demonstrated the clearest clinical value so far.
The Case for Confidence in the Numbers
Supporters of the FDA’s accelerating pace of AI device clearances argue it reflects genuine technological maturation rather than regulatory laxity, noting that the agency has simultaneously built out a dedicated Digital Health Center of Excellence specifically to evaluate AI and software-based medical devices with appropriate rigor. Proponents also point out that most cleared devices function as decision-support tools that augment rather than replace clinician judgment, preserving a human check on AI-generated recommendations even as the technology becomes more deeply embedded in clinical workflows.
The Case for Caution
Critics, including some bioethicists and patient safety researchers, argue that the sheer volume of clearances, now exceeding 1,600, has outpaced the healthcare system’s ability to monitor how these tools actually perform once deployed in real-world clinical settings rather than in the controlled conditions of a regulatory submission. There is also a transparency concern: many AI medical devices use proprietary algorithms that clinicians cannot fully inspect, raising questions about accountability when an AI-assisted diagnosis turns out to be wrong, and about whether the 510(k) substantial-equivalence pathway is rigorous enough for a technology category evolving as quickly as machine learning.
What’s Next
The FDA has signaled it will continue refining its regulatory framework specifically for AI-enabled devices, including how to handle models that continue learning and updating after their initial clearance, a scenario the agency’s traditional one-time approval process was not originally designed to accommodate. With clearances running well ahead of last year’s pace, 2026 appears likely to set a new annual record, making the question of post-market surveillance, not initial clearance volume, the more pressing regulatory challenge heading into 2027.
Photo: zillurrahmanraju / PIXABAY via Pixabay
