Decoding AI in software as a medical device (SaMD)

Use of the term artificial intelligence has become ubiquitous across multiple sectors including medical devices. Although the FDA approved the first AI-enabled medical device in 1995, the use of AI in medical devices stalled until 2015.  Since then, the number of medical devices using AI has increased rapidly with the FDA approving more than 1,000 such devices by March 2025. 

While there is not yet a unique FDA regulatory pathway, most AI-based medical devices follow a 510(k) approval pathway based on risk level and similarity to a predicate device. The remaining approvals have been granted on a De Novo basis.

This whitepaper provides a starting point for medical device researchers considering use of AI in SaMDs. It also touches on key distinctions and regulatory pathways for AI in SiMD, helping innovators plan their development of software in a medical device and software as a medical device.
Understanding FDA regulatory requirements, alongside frameworks such as IMDRF AIML and Good Machine Learning Practice for medical devices can enable medical device researchers to plan their route to market and avoid potential obstacles on the way.

Read our whitepaper to learn more about:

• AI in clinical applications, the difference between adaptive and locked AI
• The roles of AI in SaMD and AI in SiMD
• Regulatory insights, including references to IMDRF AIML and Good Machine Learning Practice (GMLP) for machine learning medical devices
• Real-world examples of AI in SaMDs, SiMDs, and broader medical device AI deployments