Anthropic proposes thresholds for halting AI models

Anthropic, a leading company in artificial intelligence, has been working on a draft bill aimed at establishing a regulatory framework for the safety of advanced AI models. According to reports from Politico, this proposal has been presented in Capitol Hill and could significantly impact how AI is regulated in the United States.

What stands out is that Anthropic's proposal sets a very high threshold for the government to intervene and halt an AI model. Specifically, it states that at least 50 deaths or economic damages exceeding $1 billion would be required before the possibility of stopping a model could be considered. This raises questions about the government's ability to respond to potential risks associated with AI.

Why it matters

Anthropic's proposal is noteworthy because it is unusual for a tech company to attempt to draft the rules for its own industry. This initiative could influence the regulatory framework for AI in the United States, affecting how these technologies are developed and used. If approved, regulation could become more lenient, which could have repercussions for public safety and social welfare.

What we know

Currently, the U.S. Congress is considering several proposals related to artificial intelligence safety. Anthropic's draft aligns with its Advanced AI Framework, which the company has been promoting since June. The intention behind this legislation is to ensure that AI models are developed responsibly and safely.

What remains unclear

It has not yet been confirmed whether Anthropic's draft bill will be taken into account by legislators or if it will undergo significant changes. Additionally, details on how the risk assessment process would be implemented and what criteria would be used to determine damages and deaths associated with AI models have not been specified.