In policy paper, OpenAI diverges from White House on AI safety

OpenAI Group PBC’s newly released proposal for how advanced artificial intelligence should be regulated differs slightly from the Trump administration’s executive order, also released this week.

Released Tuesday, OpenAI’s paper, “Democratic Governance of Frontier AI: A blueprint for a federal framework,” asks that civilian agencies be responsible for overseeing the safety of frontier AI, while the White House has placed the National Security Agency in charge of evaluating potential risks. OpenAI’s preference is that the job should fall to civilian scientific regulators at the Center for AI Standards and Innovation, or CAISI, part of the Commerce Department’s National Institute of Standards and Technology.

Nonetheless, OpenAI Chief Executive Sam Altman has voiced his support of the executive order on social media. “The U.S. should lead on AI by continuing to develop the very best models, making sure they’re safe, and getting cyber tools into the hands of trusted defenders,” he wrote in a post on X. “The new EO gets the balance right.”

Altman visited the White House today and will spoke with White House officials and key lawmakers from both the Democrat and Republican parties in what is scheduled to be a week of discussions on the future of AI. It’s reported that Altman will meet Sen. Bernie Sanders, who has been outspoken on AI and how it will affect the workforce. He will also discuss regulation with Speaker Mike Johnson (R-Louisiana) and Senate Minority Leader Chuck Schumer (D-New York).

Politico spoke with OpenAI executive Chris Lehane about the visit, saying that the company is pleased with the executive order but will attempt to push President Trump to hand over evaluation process to CAISI. He noted that the agency has the “sophisticated testing” needed for such a process and that OpenAI and its competitor Anthropic PBC have already shared information with CAISI while developing a close relationship.

Another concern of OpenAI’s according to Lehane is the White House’s plan to establish a “benchmarking” process for frontier AI models. He told Politico that this may lead to some confusion as to how much scrutiny their models will come under from the NSA.

The EO described the benchmarking process as assessing “the advanced cyber capabilities of AI models and determine the threshold at which an AI model should be designated a ‘covered frontier model’ for the purposes of this order, sharing such assessments with AI developers and researchers as appropriate.”

“I think one of the items here is, when do you hit the capability threshold?” Lehane said in the interview. “I think that’s a big part of what the conversation will be — can you establish some criteria of what that is?”

These will be some of the issues discussed in the upcoming meetings.

Photo: Wikipedia