US Eases Restrictions on Anthropic's Mythos AI Model
News Desk
28 June 2026, 17:54
The U.S. government has allowed Anthropic to restore access to its advanced Mythos 5 artificial intelligence model for a limited group of trusted U.S. organizations, partially reversing restrictions imposed earlier this month over national security concerns.
Anthropic announced on Friday that organizations responsible for operating and defending critical infrastructure will once again be able to use Mythos 5. According to a source familiar with the directive, more than 100 companies and institutions, including many Fortune 500 firms, are expected to receive access.
The move comes just two weeks after Anthropic suspended access to its most advanced AI models - Mythos 5 and Fable 5 - following new U.S. government export control measures.
In a statement, Anthropic said it is working with U.S. authorities to broaden access to Mythos 5 and eventually restore public availability of Fable 5.
Growing Government Oversight of Frontier AI
The decision reflects the U.S. government's increasingly active role in overseeing the release of advanced AI systems amid concerns that powerful models could be misused for cyberattacks or by hostile state actors.
On the same day, OpenAI confirmed that it is limiting the initial rollout of its GPT-5.6 model family to a small group of trusted partners following a request from the U.S. government. The company has said it expects a broader public release in the coming weeks.
Earlier this month, President Donald Trump signed an executive order establishing a voluntary framework under which developers of frontier AI models can provide their most advanced systems to the U.S. government for review before wider deployment.
According to Anthropic, the government informed the company that organizations approved to use Mythos 5, including eligible employees who are not U.S. citizens, will no longer require individual export licenses. However, licensing restrictions will remain in place for organizations that are not on the approved list.
A source familiar with the matter said many of the approved organizations are members of Project Glasswing, Anthropic's initiative involving around 100 technology companies and institutions.
Critics Raise Transparency Concerns
The government's role in determining which organizations receive access to advanced AI models has drawn criticism from some legal experts and industry leaders.
John Coleman, legislative counsel at the Foundation for Individual Rights and Expression, argued that the approval process lacks transparency and raises broader questions about government oversight.
OpenAI CEO Sam Altman also questioned the approach, writing on X that while extensive safety testing is reasonable, he does not support the government deciding which customers can access frontier AI models.
According to Reuters, U.S. Commerce Secretary Howard Lutnick told Anthropic that the company had made "significant progress" in addressing security concerns associated with the covered models. However, neither the government nor Anthropic disclosed the specific safeguards that led to the partial easing of restrictions.
Anthropic has previously said officials were concerned that Fable 5 could potentially be jailbroken to assist in identifying software vulnerabilities. Mythos 5 and Fable 5 are built on the same underlying AI model, but Mythos is designed for specialised enterprise and cybersecurity applications, while Fable is intended for broader public use.
Analysts say the latest decision offers a temporary path for deploying advanced AI systems but also highlights the absence of a long-term regulatory framework governing the public release of frontier AI models. As competition between U.S. and Chinese AI companies intensifies, how governments balance national security with innovation is expected to remain a central issue for the industry.
