How AI-Enabled SaMD Are Reshaping Regulatory Routes and Medicine

New AI-specific regulatory pathways are emerging for AI-enabled Software as a Medical Devices (AI SaMD), reshaping how these tools reach the market and how they are monitored in clinical practice. In the US, the FDA has established three main premarket pathways for AI SaMD:
510(k) clearance (for moderate-risk devices showing substantial equivalence to a predicate), De Novo classification (for novel, low-to-moderate-risk devices without a predicate), and Premarket Approval (PMA) for high-risk, Class III devices supported by extensive clinical evidence.

A key development is the formalization of Predetermined Change Control Plans (PCCPs), finalized in 2025, which allow manufacturers to pre‑authorize algorithm updates without filing a new 510(k) for each change, easing post-market evolution of AI models while maintaining safety oversight. In parallel, post-market surveillance requirements for AI/ML devices have been strengthened, mandating ongoing real‑world performance monitoring for many Class II AI devices.

At the EU level, AI-enabled SaMD must comply with both the Medical Device Regulation (MDR) and the EU AI Act. Until 2027, the primary route remains MDR/IVDR certification via Notified Bodies; full AI Act high‑risk obligations for most AI‑embedded medical devices apply from 2 August 2027, while certain standalone Annex III systems face these requirements from 2 August 2026. Developers must now integrate MDR/IVDR technical documentation with AI Act‑specific documentation on data governance, human oversight, robustness, and transparency.

Globally, regulators are adapting to the iterative nature of AI:
the FDA has issued updated guidance that excludes some lower‑risk digital health tools from the medical device definition, allowing faster market entry for select AI functions, while still requiring SaMD with therapeutic or diagnostic impact to undergo full premarket review. These evolving frameworks are making AI SaMD both more closely regulated and more agile, encouraging innovation while tightening oversight on safety, bias, and performance in real‑world use.