On 7 April 2025, the Food and Drug Administration (FDA) closes the public consultation on its draft Guidance for Industry and FDA Staff regarding the testing requirements for Artificial Intelligence-Enabled Device Software Functions. The draft guidance provides a series of recommendations for the validation of AI-enabled devices. These recommendations include testing methodologies, the assessment of subgroup performance, and the assurance of both safety and effectiveness. In particular, the document stipulates the necessity of validation testing to confirm the device's performance under real-world conditions, ensuring the independence of training and testing datasets. It is further recommended that testing should address AI-specific risks, including adversarial robustness, bias, and model performance drift, in order to maintain reliability and user trust.
Original source