Anthropic Proposes Federal Law to Block Risky AI Models

Anthropic is circulating a draft bill among United States lawmakers to establish a federal regulatory framework for artificial intelligence safety and security. The proposal would grant the federal government authority to block the release or use of an AI model, subject to judicial approval, if it poses an imminent risk of causing more than $1 billion in damages or killing or injuring more than 50 people.

These restrictions would apply specifically to risks involving the creation of biological weapons, the execution of cyberattacks, or models evolving beyond human control. The draft requires frontier developers to implement security programs, maintain transparency, conduct testing, and submit to independent evaluations. If passed, this federal law would override certain state-level regulations, including those in California.

The move follows recent security incidents involving autonomous AI agents and occurs while the administration of Donald Trump has relied on voluntary industry pledges. While some experts view the proposal as reasonable, others argue it is weaker than the FRONTIER Act currently in the House of Representatives. The United States Congress remains out of session until after the midterm elections.