From The Next Web · Ana-Maria Stanciuc · · 2m
Anthropic rewrites AI safety bill after safety backlash
The AI company's leaked draft drew fierce criticism from safety advocates who said it weakened protections. Anthropic says it will revise the proposal.
In brief
The AI company's leaked draft drew fierce criticism from safety advocates who said it weakened protections. Anthropic says it will revise the proposal. Anthropic will rewrite its draft AI safety bill after safety advocates said it set dangerously high thresholds for blocking dangerous models and would preempt state…
Anthropic's draft bill leaked and criticized
Politico published Anthropic's confidential AI safety bill draft on Friday. The AI safety community spent days criticizing it as weaker than existing proposals.
The bill sets high thresholds for blocking models
The draft lets government block a model only if the risk is foreseeable and imminent, involves more than one billion dollars in damage, or causes death or serious injury to more than fifty people.
Only Commerce Department can petition courts
To block a model, the Commerce Department would have to petition a federal court and make the case itself. No other agency or official could suspend a model.
The bill would preempt some state AI laws
Anthropic's draft would override state laws covering public disclosure, third-party evaluation and incident reporting. Child-safety and civil-rights laws would remain in place.
Critics compare it to the stronger FRONTIER Act
Representatives Jay Obernolte and Lori Trahan introduced the FRONTIER Act in July. Critics say Anthropic's draft sets no standards for evaluators and makes models much harder to block.
California senator says it falls short
California State Senator Scott Wiener, who wrote the state's frontier AI law, said the proposal falls far short of protecting public health and safety and sets no minimum standards.
“Anthropic's AI safety proposal falls far short of protecting public health and safety. Congress must go further.”
Startup advocates say it would hurt innovation
David Sacks, until recently the White House AI czar, said the draft's lower size threshold would sweep in most of Silicon Valley and hurt startups.
Anthropic commits to revising the draft
Anthony Cimino, leading Anthropic's federal affairs team, said the preemption wording was not precise enough and would be updated. The company will publish further proposals on independent evaluators.
“the preemption wording was "not precise enough" and would be updated.”
Anthropic faces pressure from multiple directions
The Trump administration used export controls to shut down two of Anthropic's advanced models earlier this year. Rogue-AI incidents over the summer have raised pressure for stronger rules.
The short version
Anthropic will rewrite its draft AI safety bill after safety advocates said it set dangerously high thresholds for blocking dangerous models and would preempt state laws.
