Anthropic urges Congress to keep state AI rules until federal law passes

Anthropic is urging Congress not to wipe out state AI rules unless lawmakers first enact a federal law strong enough to address catastrophic risks and require independent safety testing for its most powerful models. The statement, released Wednesday, landed as President Donald Trump continued pushing Congress to pass legislation that would supersede state rules, and as House lawmakers advanced their own bipartisan draft to override some state AI laws.

The company’s position cuts against the idea of a simple federal takeover. Anthropic is effectively asking Washington to preserve a regulatory floor until it can prove it has something better to replace it. That places the company in the middle of a broader fight over who gets to shape AI oversight first: states trying to write their own safeguards, Congress weighing national rules, and rivals that may prefer a single federal standard over a patchwork of local limits.

The timing is especially sensitive because Anthropic is preparing for a U.S. initial public offering. Investors will be watching not just its growth prospects, but also how it handles governance, safety and political risk in a business where regulation is moving from theory to active lawmaking. The company has also been under pressure from the Trump administration after refusing to let the U.S. military use its models for domestic surveillance and fully autonomous weapons systems.

That dispute has already produced concrete federal consequences. A Congressional Research Service report said Trump directed federal agencies on February 27, 2026, to stop using Anthropic technology, and Defense Secretary Pete Hegseth directed the Pentagon to designate Anthropic a supply-chain risk to national security. Reuters reported on June 5 that tensions were starting to ease in parts of the government, but the underlying conflict remains tied to how aggressively Washington wants to control advanced AI systems.

Anthropic also broadened its message beyond model safety and into labor policy, urging Congress and the states to modernize unemployment-benefit systems in case AI drives layoffs. That warning lines up with public concern: a Reuters/Ipsos poll found 53% of Americans feared AI could put someone in their household out of work, with concern spread fairly evenly across age, gender and education groups. The June 4 House draft underscored why the debate is urgent, because it would override some state laws while requiring top developers to disclose safety and security risks. Anthropic’s message was clear: do not erase state experimentation until Washington can show a credible federal framework is ready to take its place.