OpenAI backs reverse federalism to build US national AI rules

Big tech usually fights regulation until the bitter end. OpenAI is trying something different: handing lawmakers the playbook and asking them to run it.

OpenAI Chief Global Affairs Officer Chris Lehane laid out what the company calls "reverse federalism" for frontier AI safety. Since federal legislation moves slowly in Washington, OpenAI wants key states to pass aligned bills that effectively create a national standard from the ground up.

Three states, three rules

The playbook points to California, New York, and Illinois as the test cases. In OpenAI's view, California built the disclosure foundation, New York showed other states could copy the model, and Illinois added third-party verification to make those disclosures stick.

Lehane argues that state laws should strictly stick to three pillars: documented safety frameworks with public disclosures, mandatory reporting of serious incidents, and independent audits.

Anything beyond that, OpenAI warns, turns into regulatory sprawl. The company specifically told state lawmakers to stay out of national security and deep technical reviews, arguing states do not have the clearance or the compute experts to handle federal-level evaluations.

Racing toward August and beyond

At the federal level, the Trump Administration is working with technical and national security teams to finalize a cyber testing framework for frontier models by early August. OpenAI says it is actively consulting on the rollout.

On Capitol Hill, a bipartisan draft from Representatives Jay Obernolte and Lori Trahan is already circulating. OpenAI also wants lawmakers to empower the Center for AI Standards and Innovation (CAISI) to run front-end evaluations rather than just investigating harms after they happen.

The end game is global. Following G7 discussions that included non-members like India, Brazil, and Kenya, OpenAI CEO Sam Altman pitched a US-led global governance forum in the Financial Times. Google DeepMind CEO Demis Hassabis published a paper this week echoing the push for coordinated international guardrails.

Why it matters

For developers and founders, a unified state baseline is vastly better than a fifty-state compliance nightmare. If California, New York, and Illinois set the rules, most software teams only have to build their compliance pipelines once. But it also means independent audits and mandatory reporting are moving from corporate buzzwords into enforceable law.

We think OpenAI's "reverse federalism" is a clever piece of political offense. Instead of letting aggressive state lawmakers draft unpredictable restrictions, OpenAI is actively defining what reasonable oversight looks like. By pushing technical evals up to federal bodies like CAISI, it keeps state regulators from poking around in proprietary training pipelines.

The catch? A regulatory blueprint drawn by frontier giants will always tilt in their favor. By locking in heavy audit and safety reporting mandates, OpenAI is setting a bar that well-funded incumbents can clear easily, but one that could become a real barrier to entry for smaller labs trying to build frontier systems.