Anthropic has sent a draft artificial intelligence safety law to the US Congress that would allow courts to block the release or use of advanced AI models, infohub.kz reports.
Under the document, government authorities could intervene only when there is a foreseeable and imminent catastrophic risk threatening the death or serious injury of more than 50 people, or material damage exceeding $1 billion.
The company's proposal gives the US Commerce Department the exclusive right to block AI systems, through a federal judge's ruling. Other federal agencies and states would be barred from restricting neural network releases on their own.
Anthropic representatives said the document reflects the company's own vision for working with advanced AI and offers a reasonable approach to government regulation of the industry.
The draft law has sparked debate among experts and rights advocates. Critics point to the high threshold for state intervention, noting the difficulty of accurately predicting the number of potential victims before a system is launched. In addition, restrictions on the actions of individual states could weaken local AI safety legislation, making oversight excessively centralized.