Anthropic has reportedly pitched lawmakers a draft bill to stop AI models that threaten lives or cause damage
IA en un minuto newsroom · Editor: Jon Elgezabal
In 30 seconds
Anthropic has privately taken its own legislative proposal on the most powerful AI models to US lawmakers, according to Politico. The draft would only allow a model to be blocked over an imminent risk of damage above $1 billion or of more than 50 people killed or seriously injured, tied to bioweapons, cyberattacks or loss of control. Commerce would have to go to a federal judge, and no other agency could stop it.
On Friday, October 16, Politico reported that Anthropic representatives have discreetly presented their own draft bill on the safety of the most advanced AI models to members of the US Congress. The outlet cites three people familiar with those efforts.
Under that draft, the government could only block a model's launch or use if it carries a foreseeable and imminent risk of causing over $1 billion in damage, or of killing or seriously injuring over 50 people. That risk would have to come from a bioweapon being created or released, from a cyberattack or from the model escaping the control of the company that develops it.
Independent evaluators, registered with the Department of Commerce, would review the lab and its most advanced models at regular intervals. Companies could pay them, as long as the payment does not depend on the outcome of the evaluation. If they find an imminent risk of the kind the text lists, they would have to notify Commerce within 72 hours of finishing the evaluation.
After determining that such a risk exists, Commerce could ask a federal judge at the US District Court for the District of Columbia, in Washington, to order a model's suspension. The text expressly bars any other department, agency or officer of the US government from suspending its development, its deployment or its internal use.
Neither Anthropic nor the Department of Commerce immediately responded to requests for comment from Reuters, whose report on Politico's story was published by The Economic Times.
Why it matters · analysis and opinion
The draft says as much through what it allows as through what it prevents. Anthropic proposes that outside evaluators review the labs and that warnings be mandatory, but in exchange it concentrates the power to stop a model in a single department and a single court, and only for very specific catastrophic risks. Dangers that fall short of that level would not be enough to halt a model this way, and no other agency could do so on its own. It is also a text that has been shown in private, so it may change a great deal before it moves through Congress, if it ever does. For anyone using these models in their business, the signal is that the debate in Washington is no longer about whether to regulate the most advanced AI, but about who has the final say and within what limits.
Source: Politico · Written with the help of AI: how we make the news
