A harm-centered approach to regulating AI should prioritize protecting fundamental interests such as privacy, autonomy, equality, and safety regardless of the specific technology.

Develop a flexible, forward-looking regulatory framework that can address known harms and adapt to unforeseen future risks, rather than pursuing blanket regulation or minimalism.

Model regulatory architecture on established fields like environmental law, focusing on ongoing assessment, monitoring, disclosure, and mitigation instead of technology-centric rules.

Implement a practical framework with harm assessments during development, post-deployment monitoring, and reporting obligations for serious or recurring harms, ensuring notice and redress for affected individuals.

A coherent architecture is needed to connect protected interests, institutional obligations, and individual rights, addressing the current scattered and technology-specific regulation.

Recent incidents of AI-related harms—from security breaches to youth suicide linked to AI usage—underscore regulatory urgency.

While some high-risk rules will be necessary (e.g., for deepfakes or high-stakes AI decisions), the organizing principle for governance should remain harm-focused.