OpenAI plans to press Congress for mandatory safeguards on the most powerful artificial intelligence systems, which it calls "speed limits," in a policy campaign timed to begin right after Tuesday's midterm elections, Axios reported Sunday. If such rules advance, they would set binding obligations for the companies that build and deploy frontier models, the most capable general-purpose AI systems.
What Happened
Axios's Jim VandeHei reported that OpenAI's push will start during the lame-duck session of Congress, which opens Nov. 9 and runs until new members are sworn in Jan. 3. Chris Lehane, OpenAI's chief global affairs officer, told Axios his team has been building the proposals since August; Axios described them as a mix of new and repackaged ideas.
"You're going to see us talk more and more about the concept of speed limits," Lehane told Axios, adding that people "should always remain in control of this technology." OpenAI has not published the text of the proposal, and neither Axios nor TechCrunch, which followed up on the report, described how the limits would work.
The agenda has three other parts, according to Axios: community compacts and proposed state legislation giving residents a say in how data centers are built and what benefits they bring; measures to help workers through AI disruption, widen training and ease starting a business; and pressing the U.S. to help set global AI standards grounded in democratic values. OpenAI also wants its newer facilities called "AI compute campuses" rather than data centers, citing a large project in Pike County, Ohio.
Axios has a content licensing and technology agreement with OpenAI, the outlet disclosed.
Why It Matters
Mandatory pre-deployment controls on frontier models would reach beyond the labs. Boards and executives that build products on those models could see changes in access, release timing and contract terms, and policymakers would have to decide how federal rules fit with state laws. Computerworld reported in October that OpenAI has urged California to require developers to keep shutdown controls for dangerous models.
Axios expects AI companies to face more scrutiny in Washington, with Democrats likely to win the House. A Reuters analysis said Raymond James policy analyst Ed Mills sees congressional probes speeding efforts to require frontier-model safety standards, independent evaluations and incident reporting, while JPMorgan strategists said continued Republican control would likely be the most supportive federal backdrop for the AI build-out.
Lehane called the data-center debate "democracy in action," Axios reported. TechCrunch noted that Amazon and Microsoft have said they will stop using nondisclosure agreements in talks with local governments, and Reuters cited a Reuters/Ipsos poll finding that only one in three Americans approve of the fast pace of data-center construction.
What to Watch
The first test is whether OpenAI releases legislative text during the lame-duck session and what it covers: caps on deployment, approval requirements, or preemption of state laws. Lehane told Axios he expects the company's urgency on frontier safety to continue "as a new Congress ramps up."
