Trump admin considering expanding AI framework to open models

(The Daily Signal) – Trump administration officials told members of the AI industry on Tuesday that expanding the artificial intelligence vetting framework to include open-weight models is under consideration, though they are currently exempted.

Administration officials held a meeting Tuesday with AI industry executives from companies including Google, Anthropic, and OpenAI, to preview its framework for voluntary government reviews of advanced frontier AI models. President Donald Trump signed an executive order on June 3 requiring a framework for prerelease testing to address growing risks associated with new AI models.

For instance, last week, OpenAI announced that two of its most advanced models had escaped a closed testing environment and breached the AI developer platform Hugging Face. A new analysis from Hugging Face on Tuesday said two OpenAI models carried out 17,600 hacking actions on the internet between July 9 and July 13.

Media outlets have reported that open models would be exempted from the voluntary framework, but sources familiar with the meeting told the Daily Signal that while open models aren’t currently included, they could be in the future.

An open-weight AI model has trained parameters, or weights, that are released to the public, allowing others to download and customize the models.

Nvidia is a top U.S. developer of open weight models. Its CEO, Jensen Huang, met with Treasury Secretary Scott Bessent, Commerce Secretary Howard Lutnick, and Trump himself last week.

Two sources familiar with the meeting told the Daily Signal that some members of the AI industry felt little guidance was given on how the framework will work, but others say that is because this is an evolving framework that will change as technology develops.

The framework will not be shared with the public, sources said. This was first reported by Axios. AI experts on both sides of the aisle have criticized the administration for keeping the information private. National security concerns and the evolving nature of the document have been given as defenses of the secrecy.

“A secret process by which the government gets early access to cutting edge AI — what could go wrong?” asked Neil Chilson, head of AI policy at the Abundance Institute.

“There may be good reasons to classify the benchmark used to test those models,” he added. “There is no good reason to hide how the program works.”

“The White House has created a framework that shapes access to the most advanced AI models,” Rep. Lori Trahan, D-Mass., said. “It’s disappointing that they have chosen to block its public release.”

Sam Hammond, AI policy director at Foundation for American Innovation, filed a FOIA request to get a copy of the framework.

“ONCD and implementing agencies (Treasury, DHS, NIST, OSTP) do not have executive privilege,” he said. “Final agency decisions like this are textbook public records.”

Theresa Payton, chief information officer under the Bush administration, told the Daily Signal no one has a great understanding of what will put their model in the penalty box. Payton said how the administration will integrate frontier models into the framework is “to be discussed and to be determined.”

“National security is a primary concern, but trust and transparency go together,” she said, “and I believe we can actually do both. We shouldn’t have to choose between national security and transparency.