WuBlockchain
Anthropic has been pitching U.S. lawmakers on an unpublished draft bill for frontier AI safety, POLITICO reports. The proposal would let the government seek restrictions on a model's release or use only when it poses a foreseeable and imminent risk of more than $1 billion in damages or at least 50 deaths or serious injuries.
The covered risks include biological weapons, loss of developer control over a model, and cyberattacks. The draft also requires regular independent evaluations of frontier models and would place some state-level AI safety rules under a federal framework.
Why it matters
The $1 billion and 50-casualty thresholds would draw a sharp line between ordinary AI products and the small set of frontier systems Washington could actually restrain. For developers, it converts an open-ended regulatory question into a defined risk bar, which is why the proposal reads as regulatory clarity rather than a crackdown. Pulling some state rules into a federal framework also addresses the industry's biggest compliance fear: a patchwork of fifty different AI safety regimes.
Market impact
A federal framework anchored on catastrophic-risk thresholds would give large AI labs a predictable compliance path while concentrating scrutiny on bioweapon, cyber, and control-loss scenarios. Watch how lawmakers treat the state preemption element and whether the independent evaluation requirement lands as an annual cost or a competitive moat for labs with the resources to run it.
Frequently asked questions
-
What thresholds does Anthropic's draft AI bill propose?
The draft would allow government restrictions on a model only if it poses a foreseeable and imminent risk of more than $1 billion in damages or at least 50 deaths or serious injuries.
-
Which AI risks would fall under the proposed federal intervention?
Covered risks include biological weapons, loss of developer control over a model, and cyberattacks that meet the $1B or 50-casualty threshold.
-
Does the Anthropic proposal affect state-level AI laws?
Yes. The draft would place some state-level AI safety rules under a federal framework, reducing the patchwork of separate state regimes.
-
Would frontier AI models face independent evaluations under the bill?
The draft requires regular independent evaluations of frontier models as part of the federal safety framework.
-
Why is Anthropic proposing an AI safety bill now?
The proposal, reported by POLITICO, is pitched as regulatory clarity: a defined risk bar for intervention plus federal preemption of state rules gives frontier labs a predictable compliance path.
WatcherGuru
CryptoSlate
CoinDesk
CoinTelegraph