Anthropic is bringing its Mythos 5 cybersecurity AI model back online after the US government gave the green light to redeploy it to a select group of American organizations. The company announced on X that Mythos 5 will be made available to “a set of US organizations that operate and defend critical infrastructure.” Semafor has reported that number is over 100 institutions, including major corporations and government agencies.
The move ends a two-week freeze that started on June 12, when Anthropic cut off all customer access to both Mythos 5 and Fable 5. The shutdown followed a government directive demanding that the company block all foreign nationals from using its models, including those living in the US and working at Anthropic. The decision came after Amazon and other companies warned US authorities that the models could be jailbroken and misused. There were also reports that a China-linked group had gained access to Mythos 5.
The shutdown was significant because it affected not just casual users but organizations actively relying on these tools for security work. Mythos 5 is Anthropic’s most capable cybersecurity model, and pulling access without warning left many operators scrambling. Fable 5, a more public-facing version designed to bring similar capabilities to a broader audience, was caught in the same freeze.
Getting the block lifted required some serious back-and-forth between Anthropic and Washington. According to Semafor, Commerce Secretary Howard Lutnick wrote a letter to Anthropic formally approving the decision. “I have determined that appropriate safeguards are in place to permit certain trusted partners to access the Claude Mythos 5 Model,” he reportedly wrote. He also confirmed that Anthropic has committed to work with the US government on protocols, standards, and future model releases.
Lutnick’s letter reportedly credited two weeks of daily talks between Anthropic and government officials as the reason progress moved so quickly. That kind of intensive back-channel negotiation is unusual, and it points to how seriously both sides are treating the security risks around advanced AI models.
Anthropic says it is redeploying Mythos 5 quickly and working to expand access further. Fable 5 is still offline, though the company says it is in active talks with the government to restore that model too. No timeline has been given for when Fable 5 users can expect access back.
The episode highlights a tension that is becoming harder to ignore in the AI industry. Governments want to keep powerful models out of the wrong hands, but heavy-handed restrictions can disrupt legitimate users, including the very infrastructure defenders these tools are meant to help. Finding the right line between security and access is a problem that has no clean answer, and this situation shows how messy it can get in practice.
