US Government Eases Restrictions on Anthropic Claude Mythos 5 Model

A Tense Regulatory Standby Ends

The United States government has partially rolled back its recent export block on Anthropic artificial intelligence technology, permitting the company to deploy its Claude Mythos 5 model to a restricted group of domestic organizations. In a letter dated June 26, 2026, Commerce Secretary Howard Lutnick informed the San Francisco-based AI developer that it could restore access to the model for more than 100 approved entities, including security firms and critical infrastructure providers. This decision marks a significant de-escalation of a clash between the Trump administration and the startup, which began when federal officials ordered an abrupt suspension of the technology earlier in the month.

The conflict traces back to June 12, 2026, when the Commerce Department blocked access to both Claude Mythos 5 and its less powerful counterpart, Fable 5, over national security and cybersecurity concerns. The initial export ban was so sweeping that it forced Anthropic to disable access even for its own foreign national employees. Since then, the company has engaged in intensive negotiations with federal authorities to establish safeguards capable of satisfying government oversight.

The Conditions of the Compromise

In his letter to Anthropic, Secretary Lutnick acknowledged that the company had made significant progress in addressing the safety risks associated with what the government terms covered models. According to Lutnick, Anthropic has committed to collaborating with the federal government on developing standards, protocols, and release frameworks for its advanced systems. Based on these commitments, the Commerce Department determined that appropriate safeguards are now in place to allow trusted partners to utilize Claude Mythos 5.

However, this regulatory relief is highly conditional. Lutnick emphasized that the Commerce Department retains the authority to adjust the list of approved organizations at any time. He also reserved the right to reevaluate and alter the licensing requirements should national security circumstances shift. Furthermore, the newly granted permission is strictly limited to Mythos 5, leaving the less advanced Fable 5 model entirely under the export ban for the time being.

The Path Forward for Cyber Defenders

Anthropic confirmed the policy change in a public statement, identifying Mythos 5 as its strongest cybersecurity model. The company stated that it is actively working to provision the approved set of cyber defenders and infrastructure providers to restore their access as quickly as possible. While pleased with the progress, Anthropic noted that it continues to negotiate with the government to expand access to Mythos 5 and to clear Fable 5 for general use once again. Dario Amodei, the chief executive officer of Anthropic, has previously navigated a complex regulatory landscape as the Trump administration balances a generally hands-off approach to technology with sudden, national-security-driven interventions.

The resolution of this standoff occurred on the exact same day that OpenAI, one of Anthropic primary rivals, released its own highly capable model, GPT-5.6, to a group of government-approved partners. This dual development underscores the intense competitive pressure in the AI sector, even as companies face unprecedented federal scrutiny. The lack of a consistent regulatory framework continues to create friction for companies attempting to deploy cutting-edge systems while navigating shifting national security directives.

This sudden pivot from a total ban to a highly controlled release suggests that the federal government is increasingly willing to act as an active gatekeeper, deciding which specific institutions are trustworthy enough to wield the industry most potent defensive tools.