SYLEN
AboutNewsConferenceMembershipDonate

Email updates

Conference, news, and membership updates by email.

Site

  • About
  • News
  • Membership
  • Waitlist
  • Donate

Conference

  • Conference 2027
  • Call for papers

Account

  • Create account
  • Membership details

SYLEN

  • Guidelines
  • Privacy
  • Terms

© 2026 Systems Leadership and Engineering Network. sylen.org.

Membership details →
Back to news
Safety-CriticalSource: tomshardware.comOctober 18, 2026

Anthropic Proposes Federal AI Safety Framework Centered on High-Severity Casualty and Loss Thresholds

Anthropic is lobbying federal lawmakers to establish statutory limits that grant the Department of Commerce authority to block AI model deployments only when risks exceed $1 billion in damages or 50 casualties. The proposed framework relies on periodic audits by Commerce Department-registered third-party evaluators funded by model developers.

Statutory Intervention Thresholds

Anthropic is actively lobbying federal and state lawmakers for AI safety legislation that establishes high liability limits prior to regulatory intervention. Under the proposed bill, federal authorities would only possess the statutory power to block the release of an advanced AI model if an evaluation demonstrates a risk of exceeding $1 billion in total damages or causing more than 50 deaths or injuries.

The proposal codifies key aspects of Anthropic’s "Advanced AI Framework," an internal policy stance the company has advocated in legislative discussions since mid-year. By setting explicit loss metrics, the framework establishes a quantitative baseline for state-level and federal governance.

Department of Commerce Oversight Structure

The proposed legislation centralizes regulatory oversight within the federal government. Under the text of the draft bill, the Department of Commerce would effectively become the sole executive agency authorized to halt the deployment or release of an AI model that crosses the specified risk thresholds.

This centralization standardizes federal oversight within a single commercial regulatory framework, constraining pre-release enforcement powers across other federal regulatory agencies.

Third-Party Compliance Auditing

To execute continuous compliance checks, the framework mandates that AI development labs undergo periodic assessments conducted by external auditing entities.

  • Registry Requirements: Third-party evaluators must be officially registered with the Department of Commerce to conduct model assessments.
  • Audit Scope: Evaluators will assess both the underlying AI models and the internal operational compliance of the frontier labs.
  • Funding Mechanisms: The bill permits frontier labs to pay evaluators directly for testing services, under the condition that financial arrangements do not compromise evaluator independence.
Read the original article at tomshardware.com.