Anthropic Pitches Congress Its Own AI Safety Bill
The Facts
- Representatives for Anthropic have privately pitched Capitol Hill lawmakers on a draft bill to regulate frontier AI safety and security, Politico reported Friday, citing three people with knowledge of the effort and a copy of the draft.
- The draft would let the government block a system's release or use only if it poses a foreseeable, imminent risk of more than $1 billion in damages or of killing or seriously injuring more than 50 people, language that mirrors California's SB 53.
- The risk must stem from the creation or release of a biological weapon, a system evading its developer's controls or a cyberattack. Independent evaluators registered with the Commerce Department would periodically assess labs and their frontier systems under the plans.
Sources Split
LEFT
CENTER
RIGHT
+3
+1
+1
ESTAB
CRIT
CENTER
PRO
ESTAB
The Spin
Pro-industry narrative
This is a real example of frontier labs pushing decisive action instead of vague promises about safety. Narrow state preemption is a fair trade when the payoff is a nationwide standard that ensures America responsibly remains the leading global AI innovator. While this bill is only a draft in need of refinement, Anthropic deserves credit for being serious about regulation.
Industry-critical narrative
Stripping state powers by passing a federal bar so high it will hopefully never be met is self-interest masked as public duty. A threshold of mass casualties or billion-dollar damage shows exactly how valuable human life really is to AI developers, reaffirming that AI legislation should never be in the hands of frontier companies.
