The U.S. government has permitted Anthropic to re-release its advanced AI model, Mythos 5, to a select group of "trusted" U.S. organizations, partially reversing an order that had suspended access two weeks prior due to national security risks.
More than 100 companies and institutions are expected to gain access to Mythos 5, with many of them being Fortune 500 companies, according to a source familiar with the directive. The government's prior action on June 12 had led Anthropic to abruptly disable its most advanced AI models, Mythos 5 and Fable 5, for all users.
Anthropic stated that the government notified them on Friday that Mythos 5, described as their strongest cybersecurity model, can be redeployed to U.S. organizations that operate and defend critical infrastructure. The company is in the process of restoring access for these organizations and continues to collaborate with the government to expand access to Mythos 5 and to make Fable 5 available for general use again.
The U.S. Commerce Secretary Howard Lutnick indicated in a letter to Anthropic that significant progress had been made in addressing risks associated with the AI models. The letter also stated that an export license will no longer be required for Mythos 5 when provided to trusted companies and their employees, including non-U.S. citizens, and to Anthropic's non-U.S. citizen employees. However, licensing restrictions will remain for companies not on the approved list.
Concerns about powerful AI systems being misused by military intelligence in countries like China and Russia have driven the administration's assertive stance on overseeing the release of advanced AI models from companies like Anthropic and OpenAI. OpenAI recently announced it was delaying the full public launch of GPT-5.6 at the U.S. government's request, limiting its access to a small group of vetted partners whose details were shared with authorities.
The vetting process for companies granted access to Mythos has drawn criticism regarding transparency and the concentration of power. John Coleman, legislative counsel for the Foundation for Individual Rights and Expression, expressed concern that the process lacks transparency and raises questions about the rule of law. OpenAI CEO Sam Altman echoed these sentiments, stating that while safety testing is important, he dislikes the idea of the government selecting customers.
Experts have warned that Mythos models, if misused, could significantly accelerate sophisticated cyberattacks, particularly in sensitive sectors like banking. It remains unclear what specific safeguards have been implemented, though Anthropic had previously acknowledged concerns that Fable 5 could be "jailbroken" to bypass safeguards against its use in identifying software vulnerabilities.
Many of the approved companies are reportedly part of Anthropic's Project Glasswing, which includes approximately 100 tech companies and institutions. The government is also reportedly moving towards allowing the release of Fable 5, though a timeline is not yet clear. Mythos 5 and Fable 5 share the same underlying AI model, with Fable 5 designed for wider public use and Mythos 5 having some safeguards lifted.
Anthropic's relationship with the U.S. government has been strained, notably after the company refused to allow the U.S. military to use its AI models for domestic surveillance or autonomous weapons systems, which led to the government placing it on a national security blacklist. The current restrictions on Anthropic and OpenAI follow President Trump's executive order establishing a voluntary framework for AI developers to offer "covered frontier models" to the U.S. government for review before public release.
Analysts suggest that the current interim steps, while practical, do not fully resolve the broader challenge of how U.S. companies can widely release updated AI models. There is concern that delays in establishing a clear system could allow other nations, such as China, to close the gap in AI development.