Anthropic is supporting a proposed US law that would grant the government authority to suspend AI model launches if they pose serious risks. The threshold for intervention is set at systems with potential to cause at least $1 billion in damages or result in more than 50 deaths or injuries.
The legislation also mandates periodic safety assessments of company technologies conducted by independent third parties. These evaluations would focus on identifying risks related to bioweapon creation, cyberattack capabilities, and loss of model control—critical concerns as AI systems grow more powerful and autonomous.
This backing from Anthropic, a major AI developer, signals industry acknowledgment that some form of safety-focused regulation may be necessary as advanced AI systems continue to advance.