US eases AI restrictions on Anthropic's Mythos 5 for select groups

The US government has partially lifted restrictions on Anthropic’s Mythos 5 AI model. Access is now limited to a pre-approved list of organisations, while the general public remains blocked. The move follows weeks of discussions between the company and officials. On 12 June, the government imposed a universal block on Mythos 5 and its consumer-facing counterpart, Fable 5. This affected all users, including businesses relying on the enterprise model. The shutdown was abrupt and widespread.

On 26 June, Commerce Secretary Howard Lutnick sent a letter to Anthropic outlining significant progress in their cooperation. The government formalised the existing Permission Layer, turning it into a tiered access system. Mythos 5 was restored for organisations listed in Annex A of the letter, including their foreign national employees, who no longer require individual licences.

Anthropic has stated it is working to expand access to Mythos 5 for the approved providers as quickly as possible. The company also aims to restore general availability for Fable 5 in the future. Its investment in safety measures played a key role in becoming the first to have restrictions on Mythos 5 eased. Mythos 5 is now available only to vetted organisations under the new tiered system. Fable 5 remains restricted for the general public. The government’s decision reflects a priority for controlled enterprise access to frontier AI models.