OpenAI pushes AI safety through states because Washington still has no framework

OpenAI published an argument for what it calls reverse federalism: aligned state laws can help create a shared US framework for frontier AI safety. The company points to California, New York and Illinois as examples.

States are being cast as the route to national AI rules

OpenAI says the state efforts are converging around three elements: a documented safety framework with risk assessments and public disclosure, reporting of serious safety incidents, and governance through independent audits.

The primary OpenAI page was blocked during direct verification, but the text was available through a text proxy. The core claim is clear: state action can create a de facto national baseline while federal rules remain unfinished.

Safety policy is also industrial policy

The important layer is not only model safety. OpenAI argues that fragmented state regulation could slow the US ability to build a democratic AI stack and compete with actors that misuse AI in cybersecurity and other domains.

That is a disciplined political frame. The company accepts the need for rules, but steers the debate toward one baseline. For labs, that is better than 50 different regimes. For states, it preserves leverage while Washington moves slowly.

A common baseline can protect users and raise the moat

The weak point is cost. A shared safety framework can improve trust, but it can also set a threshold that smaller labs struggle to meet. Audits, incident reporting and formal risk assessments take money and staff.

The second question is definitional. If frontier model and serious safety incident remain vague, the result can become a political gesture rather than an enforceable system.

Definitions and federal adoption will decide the outcome

The next signals are whether state bills converge on precise definitions and whether the federal government turns them into a clear standard. If the overlap stays at the level of audits and safety slogans, reverse federalism will be more lobbying map than working governance.

Lilith's verdict

OpenAI is drawing a path between the lab and the legislature. Whoever holds the pen in a state bill today may guard the door to the US AI market tomorrow.