Entropic's attempt to intervene in AI legislation

Anthropic representatives have quietly submitted a bill to US lawmakers to regulate the safety and security of advanced artificial intelligence models, according to three people familiar with the matter.

Anthropic representatives have quietly submitted a bill to US lawmakers to regulate the safety and security of advanced artificial intelligence models, according to three people familiar with the matter.

According to Ambassador Arjan, quoted by Reuters, this bill allows the government to stop the supply or use of an artificial intelligence model only if it creates a foreseeable and imminent danger; A risk that causes more than one billion dollars in damages or leads to the death or injury of more than 50 people.

According to Politico, the source of this risk must be the creation or release of biological weapons, bypassing the developer's control mechanisms by the model, or the occurrence of a cyber attack.

A spokesman for Anthropic told Reuters the bill highlights proposals the company made earlier this year in its "advanced artificial intelligence framework."

The aforementioned bill allows the government to prevent or block the dangerous use of the most powerful models; It also requires developers to test leading models, make findings transparent, and accept independent evaluations.

The US Department of Commerce has not yet responded to requests for comment. According to this bill, which is registered in this ministry, independent evaluators are responsible for periodically checking artificial intelligence laboratories and their models.

The report states that laboratories can pay these evaluators, provided that the payment is not dependent on the final outcome of the evaluation. If the evaluators determine that there is an immediate risk to the company's business model, they are required to report it to the US Department of Commerce within 72 hours after the end of the evaluation process.

However, although the Ministry of Commerce can ask the judge to stop the model activity after determining the risk, the text of the bill does not explicitly allow any ministry, institution or government official to arbitrarily suspend the process of development, application or internal use of the models.