Cybersecurity Policy Report, Early Access to Frontier AI Models Would Improve States’ Cyber Defenses, Officials Say, (May 6, 2026)
Organizations Mentioned:Wintherix, LLC
As AI (artificial intelligence) developers release frontier models that can quickly identify cyber vulnerabilities, state officials are seeking early access to the models to help them improve their cyber defenses.
“State and local governments manage and oversee critical infrastructure—including power grids, water systems, and public health networks—that are more vulnerable than ever as frontier models enable bad actors to move at greater speed and scale,” according to a letter from chief information officers and security officials from 14 states.
The officials wrote to the chief executive officers of OpenAI, Anthropic, Microsoft Corp., and Google LLC to ask for “state and local government inclusion in the planning and deployment of frontier models, both generally capable models and those purposely fine-tuned for cyber capabilities.”
“While we applaud the launch of new cyber defense projects, we are deeply concerned by the current framing that prioritizes federal partners and private-sector tech giants while largely excluding the state and local entities responsible for delivering and protecting many of the nation’s most essential services,” they said.
They requested “immediate access” to models—such as Anthropic’s Claude Mythos—that have been released to a limited set of government agencies and technology firms for testing and cyber defense work (TR Daily, April 8).
“We recognize that your companies are making significant efforts to restrict harmful usage of your models. However, nefarious actors are also moving quickly to take advantage of new AI capabilities,” they noted.
“Effective cyber resilience relies on the ability to keep pace with these adversary advancements. If these powerful defensive tools remain restricted to a small circle of partners, we risk exacerbating the already asymmetric threat landscape for state and local government defenders at the front lines of our democracy,” they added.
The officials suggested holding a “follow-up roundtable within 30 days involving state leaders and your technical leadership to discuss the operational, policy, and ecosystem implications of these tools.”
The letter was signed by officials from New York, Arizona, Georgia, Hawaii, Indiana, Kansas, Maine, Maryland, Missouri, Montana, Nevada, New Hampshire, North Carolina, and Wisconsin.
News: StateLegislation DataSecurity AINews