Anthropic is pushing US states to tighten regulation of frontier AI, but it's not advocating for a one-size-fits-all approach. Instead, the company has thrown its support behind a patchwork of state-level bills: California's disclosure requirements, Illinois' annual third-party audits, and Massachusetts' stricter independent risk assessments. The Massachusetts bill would require evaluations of catastrophic risks like AI aiding the development of biological weapons, and would empower the state attorney general to enforce the rules.
OpenAI, meanwhile, is arguing for what it calls "reverse federalism" — having states adopt similar rules first, then coalescing into a national framework.
The two aren't diametrically opposed on everything. OpenAI also backed the strict Illinois bill. But with Congress dragging its feet on federal AI regulation, statehouses have become the main battleground for the two companies' competing visions.