OpenAI plans lame-duck push for mandatory AI speed limits, Axios says

OpenAI plans a policy and PR push in Washington that starts in the lame-duck session of Congress, which opens November 9, with mandatory "speed limits" on the most powerful AI systems first on its list, Axios reported on November 1. "You're going to see us talk more and more about the concept of speed limits," Chris Lehane, OpenAI's chief global affairs officer, told Axios.

The demand for mandatory rules is not new. On September 9, Lehane wrote on OpenAI's site that "the United States needs mandatory, capability-based national regulation" and that Congress should act before it adjourns. Axios itself says the speed limits build "on earlier proposals." What its report adds is the label, the timing, right after Tuesday's midterm elections, new ideas OpenAI has yet to unveil, and a wider package that ties AI safety to the fight over data centers.

What OpenAI reportedly plans

Axios described proposals, "some new and some repackaged," that Lehane's team has been stockpiling since August. They include:

• Mandatory "speed limits" on the most capable AI systems, which Lehane presented as keeping people "in control of this technology"

• New "community compacts" and commitments for places that host data centers, and state legislation giving people "a meaningful say in how projects are built and what benefits they receive"

• Support for workers, more education and training, and easier business formation

• A push for the United States to shape global AI standards built on democratic values

• Calling its data centers "AI compute campuses," with a project in Pike County, Ohio, as the example

The report did not set out how the speed limits would work: what would trigger one, who would enforce it or which systems it would cover. Axios said OpenAI "plans to unveil" them.

What OpenAI has already asked for

OpenAI's published positions give the clearest idea of what the speed limits could mean. In its September post, the company said governments should "establish shared safety bars for when and how development should slow or stop," and that "if we cannot meet certain safety bars without slowing down capability growth, we should prioritize the former."

Sources: OpenAI's frontier safety blueprint, Lehane's September 9 post and its public policy agenda, read November 1. CAISI is the federal Center for AI Standards and Innovation.

The September post also said the rules should reach only "the handful of well-resourced laboratories developing the most capable systems," not "startups, small developers, or researchers operating nowhere near the frontier." The same day, OpenAI endorsed four California bills, saying it had not backed some of them before and had reconsidered "in light of the recent jump in capabilities."

What the industry is saying

Lehane told Axios that the debate over data centers is "good and healthy" and that "communities should hold companies responsible and accountable," TechCrunch reported. He also said the midterms would be "a bit of a movie trailer" for 2028, when he expects AI to be "the issue." "I don't think it's going to be good enough just to say 'AI bad,'" he told Axios, according to AI Weekly.

TechCrunch reported that the local backlash has already moved other companies: Amazon and Microsoft have said they will stop using nondisclosure agreements in talks with local governments. SaasBuzz has covered one such local fight, the lawsuit over xAI's Memphis site.

Alex Bores, the New York Assembly member behind the state's RAISE Act, set out a test for the proposals on the morning the report ran. His question is about preemption: whether federal rules would stop states from enforcing new AI laws.

OpenAI's published answer is conditional. Its policy agenda says it supports the RAISE Act and that "with a comprehensive federal framework in place, we support the preemption of state laws that seek to regulate the same frontier safety risks." Until Congress acts, the September post says, "states should continue to move to fill the vacuum." OpenAI had not posted about the plan on its X account by 6:45 p.m. ET on November 1.

What it means for software companies

For most software companies, the rules OpenAI describes would land on their suppliers, not on them. A business building on OpenAI's or another lab's models would sit outside the "handful" of laboratories the September post names. What would reach it: OpenAI's post says it will "slow or stop the development or deployment of systems we cannot sufficiently safeguard," and asks for shared bars that would make those pauses a matter of law rather than each lab's choice. Under rules like that, release dates for the most capable models could depend partly on testing a regulator can check, which matters to any product built on them.

The data-center half of the package bears on what compute costs. OpenAI's policy agenda says it will pay "our own way on energy" and supports tariffs that make data centers pay the incremental costs they put on the grid. If the state legislation OpenAI proposes passes, host communities would gain a say in sites that labs, and the cloud providers software companies rent from, are building.

For investors, OpenAI's own documents describe the trade on offer: a binding federal framework first, then one national standard in place of state frontier laws. That sequence is what Bores is testing, and the details OpenAI brings from November 9 will show whether it holds.

What happens next

• November 3: US midterm elections

• November 9: the lame-duck session opens, and OpenAI's push begins, according to Axios

• January 3, 2027: the new Congress is sworn in, ending the session OpenAI is targeting