Anthropic Proposes Federal AI Safety Bill to Limit Government Intervention
Anthropic introduces legislation to standardize AI safety while curbing broad federal regulatory powers.
1Proposed Legislative Framework
Anthropic has submitted a draft federal bill to the US Congress aiming to regulate advanced AI models. The proposal seeks to establish national safety standards while simultaneously limiting the government's authority to unilaterally block AI releases. This initiative follows ongoing tensions between the company and federal authorities regarding AI development and deployment.
2Thresholds for Government Intervention
The bill restricts government intervention to cases involving predictable, immediate risks. Specifically, authorities could only seek to block models capable of causing over $1 billion in damages or resulting in more than 50 deaths or serious injuries. These criteria focus on catastrophic threats like biological weapon creation or major cyberattacks.
3Independent Auditing Mechanisms
Developers must implement safety systems and provide transparency reports under the proposed law. Independent auditors, registered with the Department of Commerce, would conduct periodic model evaluations. While companies may fund these audits, the compensation structure must remain independent of the results to ensure objective reporting of critical security incidents.
4Preemption of State Laws
A key component of the bill is the preemption of conflicting state-level regulations. By establishing a unified federal standard, Anthropic aims to simplify compliance for developers. This approach faces opposition from some lawmakers who fear that federal standards might weaken existing, stricter AI safety requirements currently enforced in states like California.
5Legal and Regulatory Context
The proposal emerges after a period of intense friction between Anthropic and the Trump administration. Earlier in 2026, the government blacklisted the company, citing national security concerns, though a federal judge later overturned these restrictions. This legal history underscores the company's push for clearly defined, court-backed limits on executive power.
6Centralized Judicial Oversight
The draft legislation mandates that only a federal court in Washington can issue a final order to halt an AI model's release. It explicitly prohibits other federal agencies or individual officials from independently suspending development or internal deployment. This mechanism is designed to prevent arbitrary administrative actions against AI developers.
7Competitive Legislative Landscape
Congress is currently evaluating multiple AI initiatives alongside Anthropic's proposal. The FRONTIER Act, promoted by Representatives Trahan and Obernolte, also explores state law preemption and deployment suspension powers. Meanwhile, Senate leaders are negotiating separate legislation focused on mitigating catastrophic AI risks, reflecting a broader legislative push for comprehensive federal oversight.
8Strategic Industry Implications
Anthropic’s move signals a shift toward proactive self-regulation to shape future policy. By codifying safety requirements, the company attempts to mitigate the risks of restrictive export controls and administrative blacklisting. The outcome of this bill will likely define the balance between national security interests and the operational freedom of AI firms.