FDA Proposes Updated Framework for AI/ML Medical Device Regulation

The U.S. Food and Drug Administration released draft guidance in 2023 that outlines recommendations for manufacturers of artificial intelligence and machine learning enabled software as a medical device. The document introduces Predetermined Change Control Plans to address how such devices may be updated after market entry. This step matters because AI/ML tools continue to expand in healthcare applications and require structured oversight to maintain safety standards.

Main Components of the Draft Guidance

The FDA's proposal centers on Predetermined Change Control Plans, or PCCPs, as a way for companies to specify anticipated software modifications in advance. These plans would cover changes like algorithm retraining based on new data. The guidance requires manufacturers to include protocols and assessments of potential impacts for each planned update. [1]

Handling Software Updates After Authorization

Under the framework, authorized devices could incorporate certain modifications without new full submissions if those changes fall within an approved PCCP. The approach aims to support ongoing adaptation while requiring transparency about how models evolve with real-world information. [2] The FDA has noted that traceability and oversight remain essential for any learning systems.

Anticipated Impacts on Stakeholders

Device developers would gain a clearer path for lifecycle management of their products under the proposed rules. Healthcare providers and patients could see added documentation around device performance changes over time. The FDA continues to gather public input on the draft, with the comment period closing in July 2024.

What this means

The draft guidance describes a structured method for managing post-market evolution of AI/ML software functions. It separates pre-approved change types from those that would still need separate review. Practical effects will depend on how manufacturers apply the plans and how the agency finalizes the document.

Key takeaways

• The FDA issued draft guidance on marketing submissions for AI/ML-enabled device software functions in April 2023. [1]
• PCCPs allow pre-specification of certain software updates, including algorithm retraining, to occur within defined bounds. [1]
• The framework seeks to maintain safety and effectiveness while supporting innovation through controlled adaptation. [2]
• Real-world data use in these devices requires documented protocols for transparency and oversight. [3]
• The document remains open for public comment and does not yet constitute final policy.

Limitations

The guidance exists only in draft form and may change before any final version. It applies specifically to software as a medical device and does not cover AI/ML components embedded in hardware. Actual effects on device approvals or patient outcomes cannot be assessed until manufacturers submit plans and the agency implements the approach.

Last updated: September 4, 2026