Trump administration halts AI testing unit's public reports as security concerns reshape oversight

White House Defangs AI-Testing Unit at the Worst Possible Time

There is a fight going on inside the White House over who gets to review the most powerful AI models before they are released and how much of that work the public gets to see. The Wall Street Journal reported Tuesday, citing unnamed sources, that Trump administration officials have directed the Center for AI Standards and Innovation (CAISI) to pause public reports on its AI model reviews while President Donald Trump's latest AI executive order is implemented. CAISI is part of the Commerce Department and has become the federal government's main AI testing unit. It works with companies to evaluate frontier models prior to release and shares information with the public about their capabilities. According to the Journal, the directive came amid growing concern that advanced AI models, including Anthropic's Mythos, could be misused to assist with cyberattacks or the development of biological weapons. Anthropic initially held back a full public release of Mythos, instead giving access to a limited group of companies and organizations to look for security vulnerabilities. This week, the company released a public version with additional guardrails. The news also comes as the Trump administration, which has largely given the AI industry what it wants, is trying to figure out how much oversight it is willing to impose on AI labs. So far, the administration has taken a mostly hands-off approach to regulating AI. The only time the Trump administration took any significant action against an AI company was earlier this year when it designated Anthropic as a supply chain risk after the startup refused to allow its tech to be used by the Pentagon for "any lawful purpose." Trump's latest AI executive order, signed last week after being postponed, calls for a new framework that would let AI companies voluntarily give the federal government access to frontier models up to 30 days before their wider release to "strengthen the cybersecurity of critical infrastructure." An earlier draft reportedly called for a longer 90-day review window. The Wall Street Journal says the order is a win for Treasury Secretary Scott Bessent, along with National Cyber Director Sean Cairncross, one of the officials who directed CAISI to stop publishing its reports. But not everyone in the administration appears to be on the same page. Some officials believe the executive order is assigning a new group to do work that CAISI was already doing, people familiar with the matter told The Wall Street Journal. CAISI is expected to continue evaluating models internally and working with other government agencies, but its ability to share those findings with the public now appears to be in question. According to The Wall Street Journal, top AI companies including OpenAI and Anthropic have had relationships with CAISI dating back to when the office was established during the Biden administration as the AI Safety Institute. Just last week, OpenAI called for CAISI to be strengthened. The White House and CAISI did not immediately respond to requests for comment. However, White House spokeswoman Liz Huston told The Wall Street Journal that, "The implementation of President Trump's AI agenda is a whole-of-government effort, with numerous agencies contributing to its success."

Cairncross halts AI testing unit reports on security concerns - WSJ By Investing.com

Investing.com - The Trump administration has directed a government artificial-intelligence testing unit to stop issuing public reports as the White House tightens control over AI models amid growing national-security concerns. National Cyber Director Sean Cairncross and other administration officials told the Center for AI Standards and Innovation to halt publication of its model assessments while an executive order President Trump signed last week is implemented, people familiar with the matter told the Wall Street Journal. The order represented a win for Cairncross and Treasury Secretary Scott Bessent, who have pushed for security considerations to play a bigger role in model evaluation. Housed within the Commerce Department, CAISI is the main body within the government that tests AI models before they are released and puts out public information about their capabilities and relative performance. The unit is still working internally to evaluate models and coordinate with government agencies, but stopping its public work has thrown its future in jeopardy, the people said. Top model developers including OpenAI and Anthropic have relationships with CAISI going back to the Biden administration. Companies including OpenAI have had discussions with administration officials about the importance of CAISI and preserving its power, the people said. This article was generated with the support of AI and reviewed by an editor. For more information see our T&C.