Stories
Anthropic Pitches New AI Safety Bill to US Lawmakers
- •Anthropic PBC has presented a draft bill to US lawmakers proposing new regulations for artificial intelligence safety and security.
- •The proposal suggests blocking AI models that imminently risk causing mass casualties, widespread injuries, or over $1 billion in damages.
- •Trigger conditions for blocking a model include creating biological weapons, loss of human control, or severe cyberattacks.
- •The legislation would mandate that AI developers test models, share findings publicly, and permit routine reviews by independent evaluators.
Key Questions Answered
Under Anthropic's draft bill, the federal government could block an AI model release if it imminently risks killing or injuring more than 50 people or causing over $1 billion in damages. These severe threats must stem from the model creating a biological weapon, executing a cyberattack, or escaping human control.
Anthropic's proposed legislation mandates that AI makers must rigorously test their models, share their findings publicly, and permit independent evaluators to routinely review the models. The proposal also grants the federal government powers to deter or block the deployment of the most powerful and dangerous systems.
Despite rising anxiety fueled by recent breaches, hacks, and rogue AI agents during testing, President Donald Trump's administration has so far rejected calls for implementing stricter regulatory oversight over artificial intelligence.
