Federal agencies could get modified Anthropic Mythos access under new White House plan

The US government is preparing to make a version of Anthropic’s Mythos artificial intelligence model available to major federal agencies as officials weigh the risks of using a system that has raised new cybersecurity concerns.

A memo sent by Gregory Barbaccia, the federal chief information officer at the White House Office of Management and Budget, told senior agency technology and cybersecurity officials to expect more information in the coming weeks on possible access to the model.

Bloomberg reported that the email does not confirm agencies will definitely receive Mythos access, nor does it set a specific timeline or define exactly how the model would be used. Instead, it indicates OMB is developing safeguards before any potential rollout of what Barbaccia described as a modified version of the model to agencies across the federal government.

The officials copied on the message included representatives from major departments such as Defense, Treasury, Commerce, Homeland Security, Justice, and State, underscoring how broadly Washington is already preparing for the possibility that frontier AI systems could become part of federal cyber defense workflows.

Bloomberg separately reported on April 14 that the Treasury Department had been seeking access to Mythos to help identify software vulnerabilities.

The timing is notable because Mythos is not a standard commercial launch. Anthropic introduced Claude Mythos Preview through Project Glasswing on April 7, describing it as its most capable frontier model yet for coding and agentic tasks while stressing that its unusual strength in cybersecurity is exactly why the model remains under tight controls.

Anthropic said Mythos Preview has already identified thousands of zero-day vulnerabilities across critical infrastructure and is being offered only as a gated research preview.

Under Project Glasswing, Anthropic said it is working with a limited group of launch partners to use Mythos Preview to help secure critical software and prepare institutions for a world in which frontier models can both defend and potentially endanger sensitive systems.

Anthropic also said it has been in ongoing discussions with US government officials about Mythos and its offensive and defensive cyber capabilities, adding that critical infrastructure protection is a national security priority.

That restricted rollout followed wider alarm inside both government and finance. Reuters reported on April 10 that Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell warned major bank CEOs about risks tied to Mythos after Bloomberg first reported the concerns.

Anthropic’s release of Claude Opus 4.7 earlier today adds another layer of context. In announcing Opus 4.7, Anthropic said the model is less broadly capable than Mythos Preview and is the first one being released with safeguards designed to automatically detect and block prohibited or high risk cybersecurity requests. The company said what it learns from Opus 4.7’s real world deployment will help inform any eventual broader release of Mythos class models.

That backdrop makes the government’s latest Mythos discussions especially notable. Trump directed federal agencies on Feb. 27 to stop using Anthropic’s technology, pushing parts of the government to phase out Claude and move toward rival tools from OpenAI and Google. The new Mythos discussions suggest officials may now be looking for a controlled way to bring Anthropic back into federal use, at least for cybersecurity work.