Trump White House Halts Public AI Safety Reports

Internal clash over AI oversight

The Trump administration has directed the Center for AI Standards and Innovation to suspend public reporting on its AI model evaluations, according to The Wall Street Journal. The directive affects CAISI, the Commerce Department unit that has become the federal government's primary AI testing facility, which works with companies to assess frontier models before release and shares findings with the public.

The pause comes as officials implement President Trump's latest AI executive order, signed last week, which establishes a voluntary framework allowing AI companies to provide the government access to frontier models up to 30 days before wider release. The stated goal is strengthening cybersecurity protections for critical infrastructure.

Why it matters

The decision to halt public reporting arrives precisely when transparency matters most. Advanced models like Anthropic's Mythos have raised concerns about potential misuse for cyberattacks or biological weapons development. Without public CAISI reports, the business community and security researchers lose a critical window into the capabilities and risks of models entering the market. The move also signals internal disagreement about AI oversight within an administration that has largely deferred to industry preferences.

Competing visions for testing authority

The Wall Street Journal reports that the executive order represents a win for Treasury Secretary Scott Bessent and National Cyber Director Sean Cairncross, who directed CAISI to stop publishing reports. However, other administration officials believe the order creates redundant oversight by assigning new groups to perform work CAISI already handles.

CAISI is expected to continue internal model evaluations and coordinate with other agencies, but its public transparency role now appears uncertain. An earlier draft of the executive order reportedly proposed a 90-day review window before model releases, which was shortened to 30 days in the final version.

Industry relationships at stake

Major AI companies including OpenAI and Anthropic have maintained working relationships with CAISI since its establishment during the Biden administration as the AI Safety Institute. OpenAI publicly called for strengthening CAISI just last week, before the directive to pause public reports emerged.

The Trump administration has taken a largely hands-off regulatory approach to AI, with one notable exception: designating Anthropic as a supply chain risk after the company declined to allow Pentagon use of its technology for "any lawful purpose."

White House spokeswoman Liz Huston told The Wall Street Journal that implementing the president's AI agenda is "a whole-of-government effort, with numerous agencies contributing to its success." The White House and CAISI did not respond to additional requests for comment.

These details were first reported by The Wall Street Journal.