Anthropic Advocates for Federal AI Law with Catastrophe Thresholds

Key Takeaways

1Anthropic proposes a bill to the U.S. Congress to block certain AI models that exceed risk thresholds

2The text provides for federal intervention if a model poses a risk of more than $1 billion in damages or more than 50 deaths or serious injuries

3The proposal is inspired by the California law signed in 2025 and the framework published by Anthropic in June

Why it matters : Anthropic seeks to influence AI regulation in the United States by establishing specific criteria to limit major risks associated with this technology.

Anthropic is advocating to American lawmakers for a bill aimed at suspending AI models that exceed certain thresholds of harm and casualties. The principles of a framework published in June by the company are echoed and aligned with the California law signed in 2025, according to a description shared with elected officials.

The text aligns closely with the California law signed in 2025

The bill supported by Anthropic uses language similar to that of the California AI law, signed in 2025. As early as June, Anthropic recommended adopting the definition of "catastrophic risks" used in California in its own framework of proposals. Leaders from Anthropic and OpenAI have also advocated for a slowdown in the development of cutting-edge AI models globally, to allow the industry to adapt its safeguards.

A federal blocking power conditioned on thresholds

Anthropic's proposal would give the U.S. government the ability to block the development and use of an AI model if certain risk thresholds are met. This mechanism would apply if a model poses a foreseeable and imminent risk of damage exceeding $1 billion, or if it could result in more than 50 deaths or serious injuries.

Le brief IA que lisent les pros

Le brief IA que les pros lisent chaque soir

Les 7 actus IA du jour, décryptées en 5 min. Gratuit.

Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.

Choisissez votre rythme

Gratuit · Pas de spam · Désabonnement en 1 clic

Targeted threats: biological, loss of control, cyber

The risks considered in the proposal include biological threats, situations where a model escapes the control of the company developing it, as well as cyberattacks.

A structured advocacy based on a framework published in June

Anthropic is seeking to promote its vision of AI regulation to the U.S. Congress. Representatives from the company have presented a bill to members of Congress, which is based on the proposals of a framework published in June. According to a spokesperson, Anthropic is advocating for these proposals with federal and state legislators. This framework details what the company believes governments should undertake in the short term in response to major risks associated with advanced AI, and it is available on Anthropic's website. The company develops advanced AI models, including the chatbot Claude, and regularly raises awareness about the risks associated with this technology.