US approves limited rollout of Anthropic's Mythos 5 AI model

The US government has authorised a limited release of Anthropic’s advanced artificial intelligence model, Mythos 5, allowing a small group of trusted American cybersecurity firms and infrastructure providers to access it after concerns over national security led to its suspension earlier this month.Anthropic said on Friday it had received approval to restore access to Mythos 5 for a select group of organisations. The company added that it would continue discussions with US authorities to broaden access to the model and eventually restore public availability of its less powerful Fable 5 model.The decision marks a partial reversal of restrictions imposed on June 12, when the US government ordered Anthropic to disable access to both Mythos 5 and Fable after identifying vulnerabilities in safeguards designed to prevent misuse of the technology.

Limited rollout after additional safeguards

According to CNN, US commerce secretary Howard Lutnick informed Anthropic in a letter on Friday that the company had addressed risks associated with the advanced AI models.“I have determined that appropriate safeguards are in place to permit certain trusted partners to access the Claude Mythos 5 Model,” Lutnick wrote.Anthropic confirmed the development, saying, “We received notice from the US government that Mythos 5, our strongest cybersecurity model, can be redeployed to a small group of cyber defenders and infrastructure providers.”“We are working to provision the approved set of providers and restore their access to Mythos 5 as quickly as possible. We are pleased to see this progress and continue to work with the government to expand access to Mythos 5 and make Fable 5 available for general use again,” the company added.Discussions between Anthropic and the government are expected to continue over the weekend with the aim of restoring access to Fable as well.

AI regulation under increasing scrutiny

The restrictions on Anthropic came amid growing concerns that advanced AI systems could be exploited by hackers and other malicious actors to rapidly identify and exploit cybersecurity vulnerabilitie.Earlier on Friday, rival OpenAI launched its GPT-5.6 model with access restricted on a client-by-client basis under government validation.The moves follow an executive order signed by US President Donald Trump earlier this month establishing a voluntary federal review process for advanced AI models to assess national security risks before release.However, the White House has disclosed few details about how the review system will operate or which AI models will fall under its scope.Anthropic’s relationship with the Trump administration has been strained in recent months.According to news agency AFP, the company previously opposed allowing its technology to be used for mass surveillance and autonomous weapons, a dispute that led the Pentagon to cancel contracts with Anthropic.The administration earlier designated Anthropic a “supply chain risk” over disagreements regarding military applications of its AI technology, effectively blacklisting the company. Anthropic subsequently challenged the designation in court and has secured at least one early legal victory in the ongoing case.The export restrictions and subsequent negotiations have highlighted the absence of a consistent regulatory framework for advanced AI, even as the United States seeks to maintain its technological edge over competitors such as China.