Trump Administration Plans to Move AI Testing Agency Officials to White House

The small agency within the Commerce Department tasked with testing new AI models appears to be getting a refresh.

In September, the Trump administration began the process of restructuring the Center for Advancing Innovation and Standards for Super Intelligence so that three of its senior officials report directly to the White House, according to a White House official.

For months, administration officials had considered whether to move all of CAISSI into the White House. Chief of Staff Susie Wiles and other senior officials ultimately agreed to the new arrangement in an effort to better integrate CAISSI into the White House’s decision-making process around AI, said the White House official. According to a senior Commerce official, Commerce Secretary Howard Lutnick invited two White House officials—Michael Kratsios and Sean Cairncross—to co-manage CAISSI six months ago.

The three CAISSI officials, whose names couldn’t be learned because the White House is still processing their move, will report to Kratsios, the Office of Science and Technology Policy Director, and Cairncross, Office of the National Cyber Director, the White House official said.

Still, there has been confusion between the White House and the Commerce Department on the CAISSI officials’ move, which is still being finalized. The White House official said the lack of coordination from the Commerce Department demonstrates why CAISSI leadership needs to move to the White House.

The new structure is the latest update for CAISSI, which also recently changed its name to reflect President Donald Trump’s preferred new moniker for AI. CAISSI’s new name, its third in less than two years, is a fitting symbol for an agency that has seen more than its share of tumult under Trump’s White House as AI has rocketed to the center of the national agenda.

Created by the Biden administration, the agency has been plagued by limited resources, personnel upheavals and bureaucratic nitpicking. The debate over where CAISSI should be housed has left its staff confused in recent weeks about their status, according to two people familiar with the matter.

Officials in the Trump administration have for some time debated the best place for CAISSI in the federal government. Some have argued in favor of moving the agency out of the Commerce Department in an effort to provide better resources to its staff, according to people close to the administration. OSTP director Kratsios, for one, has previously advocated for the agency to move to the Department of Energy, where CAISSI could leverage resources such as the department’s national laboratories, The Information previously reported.

The uncertainty comes at a crucial juncture for the agency and for the White House’s broader approach to AI regulation. While CAISSI has around just 40 staff and a skimpy $15 million in annual funding, it plays a key role in evaluating new AI models for their capabilities and safety.

Despite CAISSI’s diminutive footprint, its technical staff have taken on newfound importance as the Trump administration has reckoned with increasingly advanced cybermodels. As part of a White House framework finalized in August for testing models pre-deployment, CAISSI works alongside the better-resourced National Security Agency to test frontier models like OpenAI’s Astra and Anthropic’s Mythos prior to a broader release, according to four people familiar with the situation.

In a blog post published on September 30, Google announced that it had submitted its latest model, Gemini 4 Argon, to the White House for the voluntary testing process. Google still has not widely released that model.

Especially given its heightened role, CAISSI’s lack of political support and resources under the Trump administration has been a persistent source of concern for AI policy experts and lawmakers, as well as frontier labs and administration officials. When Trump announced the composition of a new SI task force in early October, it included a long list of senior administration officials. However, it didn’t include anyone from the Commerce Department and made no mention of CAISSI. U.S. Director of National Intelligence Jay Clayton, who also oversees the NSA, was named head of the task force.

Several people close to the White House and the AI industry said integrating CAISSI into the White House would be a welcome step, as it could buoy the agency’s political capital and help streamline the process for evaluating models. More clarity regarding CAISSI’s status would also address industry concerns about the White House’s approach to pre-deployment model testing, especially because it still has not released its framework, and the breakdown of responsibilities between it and other agencies like the NSA remains unclear.

Separately, Google, OpenAI and Anthropic recently explored creating their own industry group that would supplement CAISSI’s model evaluation work, partly out of concerns that the agency lacks sufficient resources, The Information previously reported.

Political Football

The Biden administration initially created the AI Safety Institute as part of a 2023 executive order on AI. It intended to staff the agency with highly technical AI experts and task it with evaluating cutting-edge new models from around the world. The institute was supposed to work hand in hand with a partner agency in the U.K., as well as with industry and policy groups, to create standards for identifying and mitigating AI risk.

The incoming Trump administration greeted the agency with skepticism. Under AI czar David Sacks, who favored a hands-off approach to regulating AI, the White House renamed it the Center for AI Standards and Innovation. The administration hired Collin Burns, a leading expert on AI testing, to lead CAISSI, but he lasted just a few days on the job. The White House fired him because he had previously worked at Anthropic, which has at times had a hostile relationship with the Trump administration, The Washington Post reported.

Though the agency remained at the Commerce Department, the administration often froze it out of broader conversations about AI, according to a former White House official. That remained true following Anthropic’s announcement of its Mythos model in April, a moment when AI’s risks became a more prominent concern within the administration and the Office of the National Cyber Director led policy discussions on the White House’s response.

The White House began to lean into CAISSI’s technical expertise more over the summer, including after the administration imposed export controls on Anthropic over security concerns with Mythos and Fable in June. And CAISSI still plays a central role in testing models as part of the White House’s AI framework, according to multiple people familiar with the process. The agency also has forged formal agreements with many of the major AI labs to evaluate their models.

CAISSI has dealt with continued turnover, however. In July, its director, Chris Fall, unexpectedly resigned after just three months on the job, and the agency is still looking for a new head, according to two people familiar with the matter. In the interim, Arvind Raman—director of the National Institute of Standards and Technology, the Commerce Department agency that houses CAISSI—is overseeing the AI agency. CAISSI’s cyber lead, Andrew Fasano, left in June to join Anthropic, according to his LinkedIn. And another key staffer, Paul Christiano, joined OpenAI’s nonprofit board in September, though his LinkedIn says he is still an adviser to CAISSI.

“President Trump is ensuring America leads the world in Super Intelligence. At the President’s request, the White House is bringing government and industry leaders together to accelerate innovation, ensure the safe and responsible deployment of new models, and protect the American people,” said White House spokesperson Liz Huston.

Leo Schwartz is a reporter covering the intersection of tech and politics. He can be reached at [email protected] or on Signal at leomschwartz.24