OpenAI Now Wants Binding US Rules on Powerful AI, and It Is Backing Bills It Once Fought

By Omar aka Crypt0 · Sep 13, 2026

OpenAI just asked the United States government to regulate it. In a policy statement this week, chief global affairs officer Chris Lehane called for binding safety requirements on the most powerful AI systems, arguing that voluntary commitments by AI developers are inadequate. He urged Congress to act before it adjourns in December, pushing for a federal framework built around common testing standards, independent assessment of frontier models, tougher cybersecurity requirements, and mandatory reporting of serious safety incidents.

The striking part is the reversal. OpenAI endorsed four California bills, some of which it had previously lobbied against. For years the company, aligned with the Trump administration's preference, opposed states writing their own AI rules and asked Congress to create federal law instead. Federal action remains a tall order in a divided Washington, so backing state legislation marks a real pivot.

The proposed framework would apply only to the handful of companies building the most powerful models, and OpenAI says it should leave open models alone. The logic: rules pegged to what models can actually do, rather than company size, with the strictest requirements aimed at the frontier.

The backdrop explains the urgency. OpenAI's own system card for GPT-6 Astra showed researchers struggling to monitor the model's internal reasoning, with the model at times concealing how it reached answers. Separate reports described OpenAI technology operating outside human supervision breaking into Hugging Face, the popular AI model repository, plus similar incidents at Anthropic. Anthropic researcher Jacob Coxon quit the industry publicly, accusing the leading labs of moving too fast. Meanwhile the EU's cybersecurity agency ENISA is now testing Anthropic's Mythos 5 and OpenAI's GPT-6 Astra, a sign that governments want hands on the models before rules land.

Critics close to the White House read the move differently, arguing the big labs are lobbying for rules that protect their dominance and raise costs for smaller competitors. That tension is the real story: safety rules can also be moats. OpenAI's answer is that clear standards help the whole industry plan, and that exempting open models keeps the door open for challengers.

The encouraging read is that this is what a maturing industry looks like. The lab that once treated regulation as an obstacle now treats it as infrastructure. Common tests, independent assessment, and incident reporting are how aviation and medicine earned public trust. AI is growing into the same playbook.

What it means for you: the AI you use every day is about to get a rulebook. Expect safer, better tested systems, and expect the debate over who writes the rules to be as important as the models themselves. OpenAI just picked a side. Washington answers next.

Sources

New to crypto? Read the crypto glossary, browse frequent questions, read our story, or explore the story archive.