What is the Mythos Preview CISA access issue?

CISA lacks access to Anthropic’s Mythos Preview before rollout

Axios reports a key access and coordination gap involving Anthropic’s Mythos Preview and the U.S. Cybersecurity and Infrastructure Security Agency (CISA). According to sources cited in the story, CISA does not have access to Mythos Preview, even though some other government agencies reportedly use it.

Anthropic, meanwhile, says it briefed CISA before Mythos’ unveiling. The dispute centers less on whether CISA received a briefing at some point and more on whether CISA has the ability to review or test the specific restricted model.

Why this matters

• Operational readiness: If CISA is expected to help with cyber risk evaluation and coordination, lack of direct access to a model that other agencies can use may slow threat assessment or response.
• Interagency inconsistency: Different access levels across government bodies can create uneven visibility into capabilities and risks.
• Security posture for restricted AI: Mythos Preview is described in the story corpus as a model withheld from public release, making oversight and evaluation especially important.

What’s known vs. unknown

We know the rough outline: CISA reportedly lacks access, and Anthropic says CISA was briefed before the model’s unveiling. What remains unclear from the summary details is how other agencies received access, what exact access mechanism was used, and whether CISA is pursuing alternative evaluation paths.

Even with Anthropic’s briefing claim, the reporting highlights a practical problem: the gap between being informed about a tool and having hands-on access to evaluate it.