The US government has allowed AI company Anthropic to restore access to its most advanced AI model, Claude Mythos 5, for a select group of more than 100 trusted companies and institutions. The decision marks a major shift just two weeks after Washington ordered the company to suspend access to the model worldwide over national security concerns.
US lifts export block on Anthropic’s Mythos 5
A letter sent by US Commerce Secretary Howard Lutnick to Anthropic on Friday, and reviewed by Reuters, said the company will no longer need an export license to export, re-export or transfer Mythos 5 to organisations listed in an attached “Annex A.” The exemption also applies to their foreign national employees as well as Anthropic’s own foreign national staff.
“Since the issuance of my June 12 letter, Anthropic has worked with the US government to address risks associated with the Covered Models,” Lutnick wrote. “These efforts have yielded significant progress.” However, the approval does not include Fable 5, Anthropic’s less powerful AI model.
A person familiar with the matter told Reuters that the approved list includes more than 100 companies and institutions, including several Fortune 500 firms. Organisations that are not on the list will continue to face restrictions.
“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,” Anthropic said in a statement.
“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 statement added.
India wants greater certainty
The sudden shutdown also worried countries that depend on advanced US-made AI models but have little control over when access can be switched on or off. India is one of them. Earlier this month, India was also included in a select group of around 15 countries that received access to Anthropic’s cybersecurity-focused Mythos models through Project Glasswing.
Electronics and IT Secretary S. Krishnan said India asked the US government for clear information about its export control policy and sought assurances that advanced AI systems would not be blocked without warning, as happened with Anthropic’s customers earlier this month.
Krishnan said Indian officials left the discussions with a better understanding of Washington’s approach. He also said they got the impression that the US plans to keep access stable for countries and partners it trusts in the future.
Because of its Pax Silica membership and earlier inclusion in trusted AI programs, Indian companies and institutions that are already on the approved list, or are added in the coming weeks, are expected to regain or continue receiving access to Mythos 5.
However, the complete list of approved organisations has not been made public, meaning some users could still face restrictions.
Anthropic Mythos: A quick turnaround after a global shutdown
The latest decision comes only two weeks after the Commerce Department ordered Anthropic to stop access to its two most advanced AI models, Mythos 5 and its safety-focused companion model, Fable 5.
The June 12 order was issued because the US government feared the technology could be used or reverse-engineered by military or intelligence agencies in China, Russia and other countries considered security risks.
The order came without warning and forced Anthropic to disable both models across the world. The company even blocked access for its own non-US employees, saying it had no practical way to limit access based only on a user’s nationality.
At the time, Anthropic strongly disagreed with the government’s decision. The company argued that the restrictions were based on a misunderstanding of a limited jailbreak technique that could easily be copied. It also said similar AI capabilities were already available through other public models that were not covered by the same export controls.