Trump Relents: US Eases Restrictions on Anthropic's Advanced AI Model

Artificial intelligence company Anthropic has reopened access to its most advanced AI model, Claude Mythos 5, after the United States government eased restrictions previously imposed on national security grounds. In an announcement on Friday (26/6), Anthropic stated the US government now permits the company to distribute Claude Mythos 5 to a number of trusted organisations within the US. The decision partially reverses a policy implemented around two weeks ago that had halted access to the model. Sources familiar with the latest policy said more than 100 companies and institutions will now gain access to Mythos 5, a list that includes many Fortune 500 member companies. President Donald Trump’s administration had adopted strict rules regarding the launch of advanced AI models from both Anthropic and OpenAI. The move was triggered by concerns that high-capability AI technology could be misused by military or intelligence agencies in China, Russia, and other countries of concern to the US government. Earlier on the same day, OpenAI also announced it was postponing the full launch of GPT-5.6 at the request of the US government. That model is temporarily only accessible to a small group of partners who have undergone a verification process and whose data is shared with government authorities. Anthropic itself had previously deactivated its most advanced AI models, Mythos 5 and Fable 5, for all users after the US government issued an export control order on 12 June. “Today, the government informed us that Mythos 5, our strongest cybersecurity model, can be used again by a number of organisations in the US that operate and protect critical infrastructure,” Anthropic said. “We are quickly restoring access for these organisations, and we continue to work with the government to expand access to Mythos 5 and to make Fable 5 available for general use again,” the company continued. US Commerce Secretary Howard Lutnick, in a letter to Anthropic, mentioned significant progress in the company’s cooperation with the government to address risks associated with Covered Models. However, it is not yet clear what additional safeguards have been implemented. Anthropic previously disclosed that the government believed there were methods to break or jailbreak the safety systems preventing Fable 5 from being used to identify software security vulnerabilities. Lutnick also conveyed that an export licence is no longer required for Mythos 5 if used by trusted companies and their non-US citizen employees or non-US citizen Anthropic staff. However, licensing restrictions remain in place for companies not on the government-approved list. Reuters sources said most of the approved companies are members of Anthropic’s Project Glasswing, which comprises around 100 prominent technology companies and institutions. The US government is also reportedly moving towards granting permission for Anthropic to release Fable 5 in the near future, although an exact schedule has not been determined. Fable 5 and Mythos 5 use the same base AI model. The difference is that Fable 5 is designed for wide public availability, whereas several safety layers on Mythos have been relaxed for specific needs. Despite this, relations between Anthropic and the US government had previously become strained. The company earlier refused to allow the US military to use their AI models for domestic surveillance or fully autonomous weapons systems. Following this, the government placed Anthropic on a national security blacklist. The restrictions on Anthropic and OpenAI were imposed after President Trump signed an executive order establishing a voluntary framework that allows AI developers to submit “covered frontier models” to the US government up to 30 days before their release to trusted partners.