Anthropic Drafts Federal AI Safety Law, Limits Intervention to Catastrophic Risk

Artificial intelligence startup Anthropic has privately circulated draft legislation to lawmakers on Capitol Hill, proposing a federal framework to regulate frontier AI safety and security. First reported by Politico, the move marks one of the most direct efforts by a leading AI laboratory to help draft the statutory rules governing advanced models, proposing a centralized enforcement structure alongside strict limits on when federal authorities can intervene.

Strict Limits on Federal Intervention

Under the draft proposal, the federal government would only have the authority to block the deployment or release of an advanced AI model under narrow, catastrophic conditions. Anthropic Combines Claude Chat and Cowork Functions into Unified Interface with New Document Tools.

Specifically, federal intervention would require an imminent, foreseeable risk of either causing more than $1 billion in damages or resulting in the death or serious injury of more than 50 people. Below those thresholds, the proposal does not grant federal agencies the power to halt model operations, establishing a deliberately high bar that excludes typical system errors and smaller-scale incidents.

Centralized Oversight Under the Commerce Department

Anthropic’s draft legislation designates the U.S. Department of Commerce as the primary regulatory channel for frontier model oversight.

Under the proposed framework, independent, registered evaluators would assess advanced models for catastrophic risks, including cyberattack capabilities, biological threat proliferation, and loss of human control. If an imminent threat is discovered, evaluators would be required to notify the Commerce Department within 72 hours. Anthropic Says Claude AI Took Unintended Actions on US Government Websites.

Rather than granting unilateral executive authority to shut down systems, the measure would require the Commerce Department to seek a federal court order to block a model's release or deployment, routing final enforcement decisions through the judiciary. The bill also incorporates federal whistleblower protections specifically designed for workers reporting AI safety hazards.

Shifting AI Regulatory Landscape

The proposal draws heavily from Anthropic’s internal Advanced AI Framework, which the company has shared with federal and state policymakers in recent months. The startup - backed by major investments from Amazon and Alphabet - has consistently branded itself as a safety-conscious alternative to competitors such as OpenAI.

By circulating ready-to-introduce legislative text, Anthropic is seeking to shape the federal standard before an increasingly fragmented patchwork of state-level regulations takes hold. While establishing binding obligations for developers across the industry, the high thresholds for intervention would prevent pre-emptive or broad regulatory restrictions on standard commercial AI deployments.

Next Steps in Congress

The draft bill has not yet been formally introduced in either the House or the Senate. Congressional staff and lawmakers are reviewing the language as debates continue on Capitol Hill over the appropriate balance between preventing national security risks and maintaining U.S. competitiveness in artificial intelligence.

(The above story first appeared on LatestLY on Oct 17, 2026 05:28 PM IST. For more news and updates on politics, world, sports, entertainment and lifestyle, log on to our website latestly.com).