Anthropic's Mythos 5 gets a limited carveout from US restrictions

Anthropic has announced it will receive a measure of relief from recent US restrictions affecting its AI models. In a statement released on Friday, the company confirmed that the government has approved limited access to its advanced model, Mythos 5. This development comes as Anthropic continues to engage with the US government since June 12, aiming to restore functionalities for both Claude Mythos 5 and Fable 5. The company emphasized that Mythos 5, their premier model for cybersecurity, will be redeployed to select US organizations responsible for critical infrastructure defense. "We are expediting access for these entities and are in ongoing discussions with the government to broaden access to Mythos 5 and eventually make Fable 5 available for broader use," Anthropic stated. The partial lifting of restrictions follows a previous order that had halted access to both AI models for foreign nationals, including those working at Anthropic's San Francisco headquarters. The initial restrictions were imposed due to government concerns that certain model safeguards might be circumvented. Anthropic has contested the urgency of these concerns, noting that the government did not clarify the specific nature of its national security worries. This situation underscores the increasing involvement of the US government in regulating the deployment of advanced AI technologies, raising pertinent questions about future access to powerful AI models. Additionally, Anthropic's competitor, OpenAI, revealed that it has also faced similar government requests, limiting preview access of its upcoming models, named GPT-5.6, to a select group of trusted partners. OpenAI voiced its concerns regarding such government interventions, asserting that they should not become a standard practice, as it restricts vital tools from users, developers, and organizations that require them for effective operations and security.