Anthropic, an artificial intelligence company, has proposed draft legislation that would allow the government to block the release of AI models only if they pose a foreseeable, imminent risk of more than $1 billion in damages or the death or serious injury of more than 50 people. The bill, outlined in Anthropic’s Advanced AI Framework and pitched to lawmakers, would concentrate regulatory power within the Commerce Department, barring other agencies from halting AI development or use. This proposal comes as lawmakers across the political spectrum debate regulating the rapidly developing AI industry, and as Anthropic increases its lobbying efforts at both the state and federal levels, spending more in the first half of 2026 than it did in all of 2025.
The legislation limits intervention-worthy “risk” to three categories: the creation of a biological weapon, the model evading developer controls, or a cyberattack, excluding other potential harms like AI-powered fraud. Anthropic CEO Dario Amodei has publicly warned of potential threats like AI-driven botnets and self-replicating AI worms, contributing to the company’s focus on existential risks, as noted in its IPO prospectus. Recent incidents, including reported breaches of testing environments at OpenAI, have heightened concerns about AI safety and control.
While lawmakers are considering various regulatory approaches, including taxes, development pauses, and new oversight departments, Anthropic’s proposal favors industry self-regulation through a single federal agency. Critics, like Miranda Bogen of the Center for Democracy and Technology, have historically questioned the effectiveness of self-regulation, citing examples like the 2007-08 financial collapse. The bill’s high threshold for intervention and concentration of power raise questions about its potential to preempt stronger state-level regulations and adequately address a broad range of AI-related risks.
Read the original coverage
💬 Comments
📜 Comment Policy