Anthropic Circulates Draft Federal AI Safety Bill Allowing Courts to Block Catastrophically Risky Models
Anthropic has been shopping a draft bill to Congress that would allow the release of an AI model to be blocked if it posed a foreseeable, imminent risk of killing over 50 people or causing over $1 billion in damage via bioweapons, loss of developer control, or cyberattacks. Independent evaluators registered with the Commerce Department, potentially paid by labs, would report risks within 72 hours, and only a D.C. judge could halt a release at the department's request. Critics note potential conflicts of interest and that the proposal stops short of giving agencies direct authority.
Skynet Chance (-0.03%): The bill explicitly targets models evading developer controls and establishes a mechanism to block catastrophic releases, modestly reducing loss-of-control risk. However, it is only a draft with weak enforcement features, such as lab-paid evaluators and no agency discretion, which limits its impact.
Skynet Date (+0 days): If adopted, pre-release evaluation and court-ordered halts could slightly slow deployment of the most dangerous models. Because it remains a proposal, the effect on the timeline is minimal.
AGI Progress (0%): The article concerns regulatory proposals rather than any change in AI capabilities, so it has no direct effect on AGI progress.
AGI Date (+0 days): The high thresholds and narrow scope suggest little friction for general development. Any slowdown in the AGI timeline would be negligible, and only if the bill passed.
<< All AI news for October 20, 2026
[ Get the daily index digest on Telegram → ]Related AI News
- Leaked Anthropic S-1 Reveals Cloud Capacity Struggles, Massive Losses, and AI Liability Risks Ahead of IPO 2026-10-20
- Anthropic Reportedly Nearing Second $45 Billion Compute Deal With Nscale 2026-10-19
- Anthropic Commits 5% of Shares to New Public Benefit Fund 2026-10-18
- Anthropic Embeds METR Evaluators With Employee-Level Access to Assess Model Risks 2026-10-18
- Anthropic Report Shows Fewer Junior Hires and More Meetings as Claude Reshapes Internal Work 2026-10-17