Cybersecurity Policy Report, AI Frontier Models Should Be Subject to Government Cyber Testing, Microsoft Says, (May 6, 2026)
Developers of frontier AI (artificial intelligence) models have an obligation to work with governments to assess the cybersecurity risks of new AI models, according to Microsoft Corp.
“While Microsoft regularly undertakes many types of AI testing on its own, testing for national security and large-scale public safety risks necessarily must be a collaborative endeavor with governments. This type of testing depends on deep technical, scientific, and national security expertise,” Natasha Crampton, Microsoft’s chief responsible AI officer, said yesterday in a blog post.
Ms. Crampton noted that Microsoft had reached an agreement with the National Institute of Standards and Technology’s Center for AI Standards and Innovation (CAISI) to test the company’s frontier AI models (CPR, May 5).
Microsoft reached a similar agreement with the United Kingdom’s AI Security Institute (AISI), she said. “In the UK, with AISI, Microsoft will collaborate on research related to frontier safety and security, including methods for evaluating high-risk capabilities and the effectiveness of the safeguards used to address them,” Ms. Crampton noted.
“These agreements matter because ongoing, rigorous testing is essential to building trust and confidence in advanced AI systems,” she said. “Testing also helps us stay ahead of risks, such as AI-driven cyberattacks and other criminal misuses of AI systems, that can emerge once advanced AI systems are deployed in the world.”
Views on whether frontier AI developers should submit their models for testing by government agencies appears to be shifting in the U.S., according to a research note published yesterday by Paul Gallant, managing director of the Washington research group–technology, media, and telecom for TD Cowen.
Faced with the development of new models that can quickly identify and exploit software vulnerabilities, the Trump administration is reportedly considering an executive order “that would require pre-release submission of additional AI models and tools that could pose cybersecurity risk to agencies and commercial users,” Mr. Gallant noted.
A similar EO issued by the Biden administration was rescinded by the Trump White House, Mr. Gallant said, but any EO that imposes testing requirements on AI developers might face legal challenges.
“In 2024 when Biden directed companies to seek govt review of ‘frontier’ AI models, his EO cited the 1950 Defense Production Act as his legal authority. But at that time, experts suggested it was far from clear that courts would accept the DPA as a legal basis for mandatory model review,” he explained.
Mr. Gallant’s view was echoed by Dean Woodley Ball, a senior fellow at the Foundation for American Innovation, and Kevin Frazier, director of the AI Innovation and Law Program at the University of Texas School of Law, who suggested that AI developers might not object to a “voluntary ‘kick the tires’ testing period.”
“Though it seems unlikely that the president can force AI labs to participate in a review system, he does not need to if the underlying goal is to translate the results of CAISI’s voluntary evaluations of model capabilities into general cyber readiness assistance. The leading labs already make their models available to CAISI for rigorous testing on a voluntary basis,” Messrs. Ball and Frazier said in a post on the Lawfare Institute blog.
“Given that they have a vested interest in avoiding the following headline: ‘AI Lab Bypasses Federal Testing; Cyberattacks Proliferate,’ they may agree for such testing to occur two or three weeks prior to generally deploying their model,” they said.
State chief information officers, meanwhile, have asked leading AI developers to also let states participate in testing (see separate story).
MainStory: TopStory FederalLegislation DataSecurity AINews