Anthropic restores access to Fable and Mythos after US lifts export ban

The United States government has lifted export restrictions on Anthropic’s most advanced artificial intelligence systems, Claude Fable 5 and Mythos 5, the company announced late Tuesday. The decision, which takes effect Wednesday, 1 July 2026, marks the end of a weeks-long standoff between the Trump administration and the San Francisco-based AI firm.

The Department of Commerce confirmed the reversal in a letter to the company, noting that a license is no longer required for the export or transfer of these models. According to the document, signed by Commerce Secretary Howard Lutnick, Anthropic has committed to proactively detecting and addressing security risks, collaborating with the government on future model standards, and reporting any instances of malicious activity.

Media additions

A Shift in Regulatory Approach

The dispute began on 12 June 2026, when federal authorities ordered Anthropic to restrict access to Fable 5 and Mythos 5 for all foreign nationals, including the company's own employees. This order was issued amid national security concerns regarding potential vulnerabilities in the models, specifically the risk that users could jailbreak or bypass internal safety guardrails to access sensitive cybersecurity capabilities. Mythos 5 is designed for business and cybersecurity experts, while Fable 5 is intended for the consumer market. Both were initially released on 9 June 2026.

While the administration initially blocked access globally, it began a partial relaxation last Friday by permitting a limited group of trusted US cybersecurity firms to access Mythos 5. Following that, the full scope of the export controls was withdrawn. The Trump administration has intensified its oversight of frontier AI models, citing fears that such technology could be exploited by military or intelligence entities in nations such as China and Russia.

Industry Reaction and Precedent

The regulatory intervention has drawn mixed responses from industry leaders. OpenAI, which recently faced pressure to stagger the launch of its own GPT-5.6 model to a vetted group of partners, has voiced concerns over the government's role in vetting users. OpenAI CEO Sam Altman remarked last week that while extensive safety testing is valuable, he does not support the idea of the government picking the customers for private technology firms. Similarly, some European lawmakers have described the US ability to cut access to such models as a kill switch that highlights the necessity for European AI sovereignty.

Francesco Bailo, deputy director of the AI, Trust and Governance Centre at the University of Sydney, suggested the initial decision to block the models likely created a messy precedent that risked significant industry backlash. Tanishq Abraham, a former research director at Stability AI, noted that the resolution of the standoff is widely viewed as a significant development, likely influenced by the efforts of Anthropic cofounder Tom Brown, who has reportedly taken a more prominent role in recent negotiations with the administration.

Context and Legal Lingering

The relationship between Anthropic and the federal government has been strained throughout the year. Earlier in 2026, the company sued the Department of Defense after the Pentagon labeled it a supply chain risk over disagreements regarding the use of AI in autonomous weapons and surveillance. While a federal judge previously blocked restrictions related to that military dispute, the government continues its appeal. Meanwhile, the Department of Commerce has indicated it reserves the right to reconsider its position if Anthropic fails to meet the newly established security requirements or if the threat landscape shifts.