The U.S. Commerce Department’s national standards agency will run tests on new artificial intelligence models from Google, Microsoft and Elon Musk’s xAI before they are released, in a step toward greater oversight of Silicon Valley by the Trump administration.
The agreement between the Commerce Department and the tech companies builds on a deal struck in 2024 by the Biden administration for prerelease testing of AI technology from start-ups OpenAI and Anthropic.
The companies have agreed to provide early access to their technology to the National Institute of Standards and Technology’s Center for AI Standards and Innovation. The center will “conduct pre-deployment evaluations and targeted research to better assess frontier AI capabilities and advance the state of AI security,” it said in an announcement Tuesday.
“Independent, rigorous measurement science is essential to understanding frontier AI and its national security implications,” CAISI Director Chris Fall said in a statement.
The agreements do not set specific standards that the companies must meet. But they mark a shift in the Trump administration’s position on AI and regulation of the industry.
At the beginning of his second term last year, President Donald Trump rolled back requirements introduced by the Biden administration that AI companies submit results of their own AI testing to the government. Trump has been a vocal proponent of giving AI companies space to innovate freely, issuing an executive order in December directing the Justice Department to sue states that pass new AI laws that the government does not approve of.
Polls have shown that Americans are increasingly skeptical of AI and largely support plans to regulate the fast-changing technology. Politicians from both main parties have railed against Trump’s friendliness with the AI industry.
The agreements to monitor new AI technology come after Anthropic shook the national security establishment last month with the announcement of Mythos, an AI model that can be used to find, fix or exploit software security flaws.
Anthropic has only provided Mythos to a select number of other technology companies, citing fears that a public release would give a powerful new tool to malicious hackers. The company’s announcement set off a scramble among government officials to try to understand exactly how capable AI models are and could be at hacking.
The agreements announced Tuesday do not appear to require companies to make changes to their technology at the government’s direction. “These agreements support information-sharing, driving voluntary product improvements and ensuring a clear understanding in government of AI capabilities and the state of international AI competition,” the Commerce Department said in its announcement.
The post Federal officials will test Google and Microsoft AI models before release appeared first on Washington Post.
