How Safe Are FDA-Cleared AI Tools?
October 1, 2025
-
3 min
7 Key Takeaways
-
1
Less than 2% of AI medical devices had randomized trial data.
-
2
Majority classified as Class II devices.
-
3
Radiology is the leading specialty for AI devices.
-
4
Significant absence of efficacy and safety reporting.
-
5
High rate of adverse events and recalls, mainly for software issues.
-
6
Calls for more rigorous regulatory measures and postmarket surveillance.
-
7
Study design details were often missing for nearly half of the devices.
A recent analysis published in JAMA Health Forum reveals that under 2% of FDA-cleared artificial intelligence medical devices included randomized clinical trial data. Examining devices from 1995 to 2023, the study found that most were Class II devices cleared through the 510(k) pathway, primarily in radiology. Key metrics on efficacy and safety were often absent, with only a small fraction reporting bias assessments or clinical outcomes. The findings highlight serious deficiencies in the regulatory and postmarket surveillance processes for AI/ML devices, calling for improvements to enhance patient safety.
Listen Tab content
