US Eases Export Restrictions on Claude Fable 5 and Mythos 5 AI Models

Key Takeaways

• The U.S. Department of Commerce has lifted export restrictions on Anthropic’s Claude Fable 5 and Mythos 5 AI models.
• The initial ban, imposed on June 12, was due to national security concerns and potential jailbreak vulnerabilities.
• Anthropic committed to new security protocols, including risk detection and reporting, to secure the unrestricted access.
• Customers will see Fable 5 access restored starting July 1, while Mythos 5 had partial reinstatement the prior week.
• Anthropic also launched Claude Sonnet 5 with an expanded context window and new pricing.

U.S. Lifts Export Ban on Anthropic’s Advanced AI Models

The U.S. Department of Commerce has officially rescinded export restrictions on Anthropic’s cutting-edge Claude Fable 5 and Mythos 5 artificial intelligence models. This decision concludes an 18-day period during which global access to these advanced systems was severely limited, impacting numerous users and researchers.

Chronology of the Restrictions and Resolution

On June 30, 2026, Commerce Secretary Howard Lutnick formally notified Tom Brown, Anthropic’s Chief Compute Officer, that an export license is “no longer required for the export, reexport, or in-country transfer, including deemed export or deemed reexport, of the Mythos or Fable models.” This communication marked the end of a contentious period for the AI developer.

The Bureau of Industry and Security (BIS) within the Department of Commerce initially mandated Anthropic to suspend access to both Fable 5 and Mythos 5 for all foreign nationals on June 12. This directive stemmed from national security concerns and perceived diversion risks. Amazon executives and National Security Agency (NSA) reviewers had reportedly highlighted potential jailbreak vulnerabilities within the models, prompting the government’s intervention.

The immediate consequence of the BIS order was a complete shutdown of both models for all Anthropic customers, regardless of their location, to ensure full compliance. This blanket ban drew significant criticism, with over 100 cybersecurity researchers publicly arguing that the restrictions impeded legitimate defensive security research efforts.

A partial easing of the restrictions occurred on June 26. Secretary Lutnick authorized limited access to Mythos 5 for a select group of government-approved organizations and trusted partners. This decision followed a BIS assessment that “appropriate safeguards” could be implemented for these specific users.

Anthropic’s Commitments Pave the Way for Full Access

The decisive June 30 letter cited Anthropic’s commitments to enhanced coordination as the primary reason for lifting the controls. These agreements include Anthropic pledging to proactively detect and report security risks associated with its models, collaborate with the U.S. government on release protocols for both current and future AI models, and disclose any malicious activities involving Fable or Mythos.

BIS’s revised evaluation of the diversion risk for both models ultimately justified the complete withdrawal of the prior restrictions. However, the Department of Commerce explicitly stated its right to reimpose licensing requirements should Anthropic fail to uphold its new obligations or if circumstances change.

Anthropic confirmed through a post on X that it would commence restoring customer access to Fable 5 on Wednesday, July 1. Mythos 5 had already seen partial reinstatement to certain users in the preceding week.

New Model Launch Coincides with Resolution

Amidst the resolution of the export control issue, Anthropic also announced the launch of Claude Sonnet 5. This new model boasts a 1 million token context window and introduces more affordable pricing tiers, building upon the context-window expansion architecture previously featured in earlier Sonnet models.

Disclaimer: HackersRadar reports on cybersecurity threats and incidents for informational and awareness purposes only. We do not engage in hacking activities, data exfiltration, or the hosting or distribution of stolen or leaked information. All content is based on publicly available sources.