The US government has partially reversed its June 12 export control order, allowing Anthropic to restore Claude Mythos 5 access for over 100 trusted US organisations including Fortune 500 companies.
What Happened: The June 12 Export Control Order
On June 12, the US government issued an export control order that forced Anthropic to abruptly disable two of its most powerful AI models — Mythos 5 and Fable 5 — for all users worldwide. The move was driven by concerns that advanced AI systems could be misused by military or intelligence actors in countries such as China and Russia.
The restrictions were part of a broader, aggressive oversight approach by President Donald Trump’s administration toward frontier AI models developed by Anthropic and rival OpenAI.
Access Now Restored for Critical Infrastructure Organisations
On Friday, Anthropic announced the US government had notified the company that Mythos 5 could be redeployed to a specific set of organisations that operate and defend critical infrastructure across the United States.
In a statement, Anthropic confirmed that Mythos 5, described as their strongest cybersecurity model, can now be redeployed to qualifying US organisations.
More than 100 companies and institutions are included in this restored access, with many being Fortune 500 firms, according to a source familiar with the directive who requested anonymity due to the sensitivity of the matter. A significant number of these organisations are part of Anthropic’s Project Glasswing, a programme that includes roughly 100 well-known technology companies and institutions.
Anthropic added that it is continuing to work with the government to expand Mythos 5 access further and restore public availability of Fable 5, though no timeline has been confirmed for a broader rollout.
Mythos 5 vs Fable 5: Key Differences
Both Mythos 5 and Fable 5 are built on the same underlying AI model, but they serve different purposes and carry different access controls.
| Model | Intended Use | Current Status |
|---|---|---|
| Claude Mythos 5 | Critical infrastructure defence, cybersecurity | Restored for 100+ approved US organisations |
| Claude Fable 5 | General public use | Still restricted; timeline for return unclear |
Experts have noted that Mythos-class models, if accessed by the wrong parties, could significantly accelerate sophisticated cyberattacks — particularly in sectors like banking that rely on complex, older technology systems.
Earlier this month, Anthropic acknowledged that the government believed a method existed to bypass, or “jailbreak,” a safeguard meant to prevent Fable 5 from being used to identify software vulnerabilities.
Licensing Rules and Non-US Citizens
US Commerce Secretary Howard Lutnick wrote a letter to Anthropic citing significant progress in the company’s work with the government to address risks tied to the restricted models. Under the updated directive:
- An export licence will no longer be required for Mythos 5 to be used by trusted companies and their employees who are not US citizens.
- Anthropic’s own non-US citizen employees will also be covered under the eased restrictions.
- Licensing restrictions remain in place for any organisation not on the approved list.
The letter did not detail the specific safeguards now in place, and the full conditions of the agreement between Anthropic and the government have not been made public.
Criticism Over Government’s Selection Process
The government’s role in choosing which organisations gain access to Mythos 5 has attracted strong criticism from civil liberties advocates and industry figures alike.
John Coleman, legislative counsel at the Foundation for Individual Rights and Expression — a nonpartisan free speech organisation based in Philadelphia — raised concerns about transparency and fairness in the process.
- He questioned how selected companies are chosen and why others are excluded.
- He argued the process gives the government too much power with too little accountability.
- He said the arrangement raises serious questions about the rule of law.
OpenAI chief executive Sam Altman also weighed in on the matter via a post on X, stating that while safety testing is not a bad idea, he objected to the idea of the government deciding which customers get access to top AI models.
OpenAI separately confirmed it was delaying the full public launch of its own GPT-5.6 model at the US government’s request, restricting access to a small group of vetted partners whose details were shared with authorities.
Broader Implications for the US AI Industry
The Trump administration’s approach to frontier AI oversight was formalised through an executive order signed this month. The order establishes a voluntary framework under which AI developers must offer their most capable “covered frontier models” to the US government for review for up to 30 days before releasing them to trusted partners.
Analysts have warned that prolonged access restrictions could hand a competitive advantage to China.
Kate Koren, an analyst at the Center for Strategic and International Studies and a former Commerce Department official, described the current directive as a practical interim step, but cautioned that without a broader system allowing companies to widely release updated models, China could close the gap with US AI developers.
Anthropic’s relationship with the federal government has also been strained by an earlier dispute in which the company refused to allow the US military to use its AI models for domestic surveillance and fully autonomous weapons systems. The government responded by placing Anthropic on a national security blacklist.
Both Anthropic and OpenAI have indicated plans to go public in the future, adding further complexity to their ongoing negotiations with Washington over access, oversight, and commercial growth.
Frequently Asked Questions
laude Mythos 5 is Anthropic's most advanced AI model, designed primarily for cybersecurity applications and the defence of critical infrastructure. It shares the same underlying model as Fable 5 but operates with different access controls and fewer public-facing safeguards.
The US government issued an export control order on June 12, citing national security concerns that powerful AI models could be misused by military or intelligence operatives in adversarial nations such as China and Russia. Anthropic was required to disable both Mythos 5 and Fable 5 for all users as a result.
Project Glasswing is an Anthropic initiative that includes approximately 100 well-known technology companies and institutions. Many of the organisations now approved to access Mythos 5 are part of this programme, which operates under close coordination with the US government.
