Trump administration to ask US AI firms to voluntarily submit models for cyber security tests

The Trump administration will ask leading ⁠AI developers ⁠to voluntarily submit their most capable models for government cyber security tests before releasing them to the public, according to an executive order, as security fears mount in Washington over powerful new AI systems such as Anthropic's Mythos.

US President Donald Trump signed ‌an executive order that directs the departments of Treasury, Defense, Commerce ‌and ‌Homeland Security, plus other government officials and agencies, to secure agreements ‌with AI developers to test their models.

US agencies would ⁠get up to 30 days to test the models before they are released to organisations outside the government, according to the order. 

It also directs the agencies to emphasize bolstering cyber defence across government.

The order signals Trump is shifting his strategy ​on AI and taking a more active role in monitoring the technology's capabilities.

Since returning to office, he has said the US federal government should take a hands-off ⁠approach to the tech sector, and has tried to discourage states from adopting AI regulations that he opposes.

The president's decision to implement voluntary testing could hurt the industry's profits if it slows the rollout of new models or prompts the companies to change how they perform to address security concerns.

Anthropic, OpenAI and Alphabet's Google met with the US government about cyber security, a senior US official said in May on a call with reporters about the executive order's development.

Google executive Kent Walker called the executive order "an important step forward." Anthropic and OpenAI ​did not immediately respond to a request for comment ⁠from Reuters.

Trump was slated to sign an executive order on artificial ⁠intelligence on May 21. He postponed the signing the same day, saying he did not like certain aspects of the order ​and did not want to take any steps that might undermine the US position in its AI ‌competition with China.

Protecting vital US sectors

The Treasury Department consulted with banks when developing the order, Treasury Secretary Scott Bessent said in mid-May.

The order asks Bessent to work with AI developers and critical infrastructure providers to scan software for cyber security flaws and ‌develop patches for those vulnerabilities.

Critical infrastructure refers to sectors that are vital to the US economy and deserving of special protection from attacks, such as banks, emergency services and hospitals. 

Voluntary federal testing has been in place for a few years, with companies such as ​OpenAI and Anthropic submitting their models for scrutiny by the US Department of Commerce's Center for AI Standards and Innovation, known by a different name under former President Joe Biden.

The department announced in May that Google, xAI and Microsoft ‌had agreed ⁠to submit their AI models for security ​testing, though the details later disappeared from its website.