Anthropic’s Claude Mythos 5 is back with limited access, only for a set of U.S. organizations. June 26, 2026, Anthropic said the U.S. government had allowed them to redeploy the model to approved organizations after a two-week suspension, citing national security concerns. Now more than 100 companies and institutions will have access, including many Fortune 500 firms and critical-infrastructure organizations connected to Anthropic’s Project Glasswing program.
Mythos 5 is Anthropic’s strongest cybersecurity model, and after the US, the company quickly moved to restore access for those organizations. The company said they are in continuous talks with the government on wider access.
The model is not back for everyone, it is back for a vetted group, with access still controlled by the government approval process and Anthropic’s trusted-access framework.
[ALSO READ: Before the Public Sees Them, the U.S. Government Will Test Top AI Models ]
A timeline
June 9, 2026
Anthropic launched Claude Fable 5 for general use and Claude Mythos 5 for a small group of vetted cyberdefenders and infrastructure providers. Anthropic said Mythos 5 was the same underlying model as Fable 5, but with safeguards lifted in some areas, and that it had the strongest cybersecurity capabilities of any model in the world.
June 12, 2026
Anthropic said the U.S. government issued an export-control directive that suspended access to Fable 5 and Mythos 5 for any foreign national, including foreign national Anthropic employees. Anthropic said the effect of the order was that it had to disable both models for all customers to stay compliant. The company also said the government had not given exact details, but Anthropic understood officials believed there may be a way to bypass, or “jailbreak,” Fable 5 and use it to find vulnerabilities.
June 26, 2026
Anthropic said the government notified it that Mythos 5 could be redeployed to U.S. organizations that operate and defend critical infrastructure. Anthropic said it was restoring access quickly and continuing to work with the government to widen access and eventually make Fable 5 available for general use again.
Who will get access to Mythos 5
More than 100 organizations will now have access, including many Fortune 500 companies and institutions tied to critical infrastructure. The report also said many of those approved groups are part of Project Glasswing, Anthropic’s restricted cybersecurity program. Anthropic’s own Mythos page says the model is restricted to Glasswing partners for now, with wider trusted access still to come.
Anthropic’s Glasswing page explains the type of work those partners do: local vulnerability detection, black-box testing of binaries, securing endpoints and penetration testing. That is the practical use case for Mythos 5 right now.
[ALSO READ: Anthropic Launches ‘Project Glasswing’ with Tech Giants to Boost Global Cybersecurity ]
What the U.S. was concerned about
Anthropic said the government’s June 12 directive was based on national-security authorities and a belief that Fable 5 might be vulnerable to a bypass or jailbreak. Anthropic said it had reviewed a demonstration of the technique and found it was used to identify a small number of previously known, minor vulnerabilities. Reuters also reported that U.S. officials were worried advanced frontier models could be misused by military or intelligence users in countries such as China or Russia.
That explains why the government did not simply allow a broad relaunch. It opted for a controlled return to organizations that Anthropic and the U.S. government consider trusted.
What Mythos 5 is able to do
Anthropic describes Mythos 5 as its most capable model for cybersecurity and biology research. The company says it is state-of-the-art in cybersecurity, biology research and healthcare, and that the model’s capabilities are strong enough that it could be used both for good and for harm.
Anthropic Mythos 5 is the same underlying model as Fable 5, but with cyber safeguards lifted for approved users. Fable 5 is the general-use version. Mythos 5 is the restricted version for cyber and biology research work.
What does limited access mean
Limited access means Mythos 5 is no longer available as a public model. It is available only to approved U.S. organizations.
Export licenses will no longer be needed for approved companies’ non-U.S.-citizen employees or Anthropic’s non-U.S.-citizen employees, but the restrictions remain for organizations that are not on the approved list.
Anthropic is working toward broader access, but for now, the model sits inside a permissions system rather than a public rollout.
The government’s selection process has drawn criticism. John Coleman of the Foundation for Personal Rights and Expression says no one knows how companies are chosen or why others are excluded, and that the process gives too much power to the government.
OpenAI chief executive Sam Altman echoed the concern, saying extensive safety testing is not a bad idea, but he did not like the government picking customers.
The question now is not only whether frontier AI should be tested, but who gets to decide which customers can use it first.
What OpenAI said about GPT-5.6
OpenAI delayed a full public launch of GPT-5.6 at the U.S. government’s request and limited initial access to a small group of vetted partners whose details were shared with officials. OpenAI said the step was temporary while it worked with Washington on a wider structure for future releases. Sam Altman said he supports safety testing, but not the government choosing customers.
That puts Anthropic and OpenAI in the same policy lane: frontier models can still launch, but not always on the companies’ preferred timetable or to the full market at once.
The limited return of Mythos 5 is a sign of how frontier AI is now being handled in the U.S. The model is powerful enough to help defenders probe critical systems, but that same capability is what triggered the first suspension.
[ALSO READ: India Opens Insurance to Full Foreign Ownership, LIC Remains Under a 20% Cap ]
