Trump Orders Voluntary AI Security Testing Plan

The Trump administration will ask leading artificial intelligence developers to voluntarily submit their most advanced models for U.S. government cybersecurity testing before public release, according to an executive order signed on Tuesday.

The move comes as Washington steps up scrutiny of advanced AI systems amid rising national security concerns over powerful models, including emerging systems such as Mythos developed by Anthropic.

Trump Expands Federal AI Testing Oversight

U.S. President Donald Trump signed an executive order directing the departments of Treasury, Defense, Commerce and Homeland Security, along with other federal agencies, to secure voluntary agreements with AI developers for cybersecurity testing before public release.

Under the directive, U.S. agencies will have up to 30 days to evaluate leading AI models before they are deployed outside government systems. The order also places emphasis on strengthening cyber defenses across federal infrastructure as AI capabilities expand.

The Treasury Department is expected to coordinate with AI developers and critical infrastructure operators, including banks, hospitals and emergency services, to identify software vulnerabilities and develop security patches.

AI Industry Engagement and Policy Shift

The executive order reflects a broader shift toward more active federal oversight of frontier AI systems, particularly in cybersecurity, after earlier signals of a lighter regulatory approach from the Trump administration.

Since returning to office, Trump has generally promoted a less interventionist stance toward technology regulation, while also discouraging state-level AI rules he views as restrictive.

However, the new policy indicates growing concern within the administration over how advanced AI systems could be exploited for cyberattacks or other forms of misuse.

Voluntary government-led AI testing is not new. Companies such as OpenAI and Anthropic have previously submitted models for evaluation under U.S. Department of Commerce-led programmes focused on AI safety and standards.

According to a senior U.S. official, firms including Alphabet Inc., OpenAI and Anthropic have participated in discussions with U.S. agencies on cybersecurity testing and model evaluation frameworks.

Google executive Kent Walker described the executive order as “an important step forward,” while Anthropic said it looked forward to working with the White House on implementation. OpenAI CEO Sam Altman said the directive “gets the balance right,” adding that the U.S. should continue leading in AI while ensuring models are safe and usable by trusted defenders.

Commerce Department Removes AI Testing Disclosure

The U.S. Commerce Department has removed website details outlining its agreement with Google, xAI and Microsoft to test artificial intelligence models for security vulnerabilities, according to a Reuters review of the agency’s website.

A link that previously directed users to the announcement is no longer accessible and now displays a “Sorry, we cannot find that page” message. The page later redirected to the Center for AI Standards and Innovation, the federal body responsible for conducting the tests.

The Commerce Department had announced on May 5 that the companies would provide new AI models prior to public deployment, allowing government scientists to assess them for cybersecurity risks. Officials said the aim was to identify potential threats ranging from cyberattacks to military misuse.

It remains unclear why the webpage was removed. Spokespeople for both the Commerce Department and the White House did not immediately respond to requests for comment.

Cybersecurity Concerns Drive Early Access Model Testing

The policy shift comes amid growing concern in Washington over the national security risks posed by rapidly advancing AI systems, including concerns that frontier models could be misused for cyberattacks or other sensitive applications.

By granting government agencies early access to advanced AI models, officials say they aim to detect vulnerabilities before systems are widely deployed, reducing the risk of exploitation across critical infrastructure sectors.

The updated framework signals an evolving approach in Washington: balancing voluntary cooperation with industry against increasing pressure to ensure AI systems are secure before they reach the public.

The executive order underscores how AI governance in the United States is increasingly being shaped by cybersecurity priorities, as regulators and companies navigate the risks of deploying increasingly powerful systems at scale.

Also Read: CcHUB expands Lagos Headquarters, introduces private startup offices in 15 years

Segun Lasisi

A Lagos-based journalist with a passion for disseminating factual information and a deep appreciation for good music, good food, movies, and beautiful cars. He hopes to travel the world someday, documenting its beauty and diverse cultures through his storytelling.