The American Medical Association urged a risk-based approach to generative AI oversight, calling on the FDA to clarify responsibility for safety once these tools enter clinical practice. The position reflects mounting pressure to define accountability boundaries for AI-driven decision support and clinically consequential automation. Separately, a report summarized FDA’s finalized guidance on drug-related apps, a regulatory signal for digital therapeutics, companion software, and how firms structure compliance for software that touches medicines. The same policy ecosystem is also drawing scrutiny in how evidence should be built for generative AI medical devices. In practice, these updates converge on a single operational challenge for developers and sponsors: mapping intended clinical function to the level of evidence and oversight needed, while anticipating how “risk” will be interpreted across real-world deployments.