Just two weeks after forcing Anthropic to pull its most advanced artificial intelligence models, the U.S. government has eased some of the restrictions, allowing a select group of organizations to use one of the company’s flagship AI systems again.
The decision marks a significant shift in the government’s approach to regulating frontier AI models, although public access remains off the table for now.
According to Reuters, Anthropic confirmed on Friday that it had been authorized to restore access to Claude Mythos 5 for more than 100 approved U.S. organizations, including several Fortune 500 companies and institutions responsible for protecting critical infrastructure.
The company said it has already begun restoring access and will continue working with U.S. officials to broaden availability in the future.
Anthropic begins restoring Mythos 5
In a statement, Anthropic said the government informed the company that Mythos 5 could once again be deployed to organizations that operate and defend critical infrastructure.
“Today, the government notified us that Mythos 5, our strongest cybersecurity model, can be redeployed to a set of U.S. organizations that operate and defend critical infrastructure,” the company said.
“We’re restoring access for these organizations quickly, and we’re continuing to work with the government to expand access to Mythos 5 and make Fable 5 available for general use again.”
While Mythos 5 is gradually returning for approved organizations, Anthropic’s consumer-focused model, Fable 5, remains unavailable to the general public.
Why the models were restricted
The partial reversal comes after the U.S. government imposed export controls earlier this month, citing concerns that Anthropic’s most advanced AI models possessed cybersecurity capabilities that could pose national security risks if widely available.
The restrictions forced Anthropic to restrict access to both Mythos 5 and Fable 5 worldwide while discussions with federal officials continued.
According to Reuters, U.S. Commerce Secretary Howard Lutnick informed Anthropic that the company had made “significant progress” in addressing concerns surrounding what officials described as “Covered Models.” However, the government has not publicly disclosed the specific safeguards Anthropic agreed to implement.
AI regulation remains under scrutiny
Although Anthropic welcomed the latest decision, the government’s approach has continued to attract criticism from legal experts and technology leaders.
Critics argue that there is little transparency in how companies are selected for access to advanced AI systems, raising concerns about fairness and government oversight.
John Coleman, legislative counsel at the Foundation for Individual Rights and Expression, questioned the process, saying there is still no clear explanation of how approved organizations are chosen or why others remain excluded.
OpenAI CEO Sam Altman has also expressed reservations about the government’s evolving AI policies.
While saying he supports rigorous safety testing for powerful AI models, Altman argued that governments should not decide which companies are allowed to access frontier AI systems.
OpenAI facing similar restrictions
The development comes as OpenAI also navigates tighter government oversight.
The company recently confirmed it would delay the broader rollout of its upcoming GPT-5.6 model after receiving a request from the U.S. government to limit access to government-approved partners while safety evaluations continue initially.
The parallel cases involving Anthropic and OpenAI suggest the U.S. government is taking a more active role in determining how next-generation AI models are released, particularly those with advanced cybersecurity capabilities.
For now, Anthropic has secured a partial victory by returning Mythos 5 to trusted organizations. Still, broader public access to its most powerful AI technology remains uncertain as negotiations with regulators continue.
