A draft of a federal AI safety bill that Anthropic has been circulating among lawmakers leaked late last week, shifting the debate over what an acceptable federal framework should look like. The document represents a legislative version of the AI safety framework Anthropic has promoted for months, a company spokesperson told Semafor.
AI safety advocates responded critically, saying the proposal does not go far enough to manage potential harms, would unfairly pre-empt stricter state laws, and accepts too much possible risk.
Anthropic signals revisions
After the leak, Anthropic — long perceived as a lab that emphasizes safety — found itself on the defensive. One of the company’s top lobbyists posted that the firm plans to make tweaks to the draft so it better reflects the rapid pace of AI progress and the company’s position on pre-empting state-level laws.
The controversy centers in part on how much a federal bill should override state authority: critics fear the draft grants developers too much leeway and constrains states from enacting more stringent rules.
OpenAI seeks to differentiate itself
OpenAI quickly seized the moment to set itself apart. Chris Lehane, OpenAI’s Chief Global Affairs Officer, wrote in a Monday post that policy decisions must ultimately be made through democratic processes led by elected officials. He added that no single company — including OpenAI — should write the rules alone for a technology with profound implications for the economy, society, and national security.
Different regulatory strategies among companies
In recent years Anthropic has positioned itself as more openly in favor of regulation than some peers. As the risks posed by advanced AI have become clearer — and as tensions with the U.S. government have grown — Anthropic has pushed for regulatory mechanisms that have provoked heated debate in federal circles, including proposals that would allow officials to block the release of certain models.
By contrast, OpenAI has generally acted as a more seasoned Washington player, aiming to be seen as willing to endorse compromise federal bills written by lawmakers while promoting similar acceptable frameworks across states.
Why this matters
The leaked draft highlights ongoing tensions between corporate influence and public interest in shaping AI law. The outcome of these debates will affect what tools and authorities federal and state regulators have to mitigate the risks of advanced models and how much influence private companies can exert over the content of those rules.
Following the leak, further edits to the draft and intensified political debate are likely as companies, legislators, and advocacy groups negotiate the next steps in the regulatory process.
