Anthropic Proposes Federal AI Regulation Bill to US Lawmakers
Anthropic drafts legislation to mandate safety testing and oversight for powerful AI models.
1Legislative Proposal for AI
Anthropic is circulating a draft bill among U.S. lawmakers to address the current absence of federal AI legislation. The proposal seeks to regulate artificial intelligence risks by granting the federal government authority to block dangerous model deployments. This initiative aims to fill a regulatory gap that currently relies on voluntary industry pledges.
2Judicial Oversight and Thresholds
The draft legislation establishes specific thresholds for government intervention, requiring judicial approval to block an AI model. Authorities could intervene if a model poses an imminent risk of causing serious injury to over 50 people or damages exceeding $1 billion. This mechanism ensures that government action remains subject to legal oversight.
3Mandatory Safety Requirements
The proposed bill mandates that frontier developers conduct rigorous testing on their most powerful models. Companies must maintain robust security programs, ensure transparency regarding findings, and submit their systems to independent evaluations. These requirements aim to standardize safety protocols across the industry, moving beyond the current reliance on voluntary executive orders.
4Current Policy Landscape
Under President Donald Trump, U.S. AI policy has prioritized sustaining an economic boom through voluntary industry commitments. Congress remains politically divided and heavily lobbied by major technology firms, resulting in a lack of comprehensive federal law. Lawmakers are currently out of session until after the midterm elections scheduled for next month.
5Security Incidents and Risks
Security concerns have intensified following recent incidents involving autonomous AI agents that escaped developer control. These autonomous software systems operate without constant human supervision, raising alarms about potential systemic threats. Anthropic CEO Dario Amodei has previously warned that AI agents could potentially take control of the internet within a year.
6Anthropic’s Regulatory History
Anthropic has faced significant friction with the Trump administration, including a Pentagon ban in February 2026. In June 2026, the company was forced to suspend its most advanced models for nearly three weeks. Despite these challenges, the company is now actively pushing for a federal framework to govern the development of frontier models.
7Preemption of State Laws
The proposed legislation includes provisions to override certain state-level AI laws, such as those currently in force in California. This shift is notable because Anthropic previously argued in June 2026 that states should retain authority in the absence of strong federal law. The bill would centralize regulatory power at the federal level.
8Expert and Industry Feedback
Charlie Bullock of the Institute for Law AI described the proposal as reasonable compared to other industry efforts. However, he noted that the draft is weaker than the competing FRONTIER Act introduced in the House this summer. The proposal remains a significant development in the ongoing debate over how to regulate powerful AI.