Anthropic Export Controls Lifted, Fable 5 Restored

Anthropic PBC said on June 30, 2026, that the U.S. Department of Commerce had lifted export controls on its Claude Fable 5 and Mythos 5 models. This decision allows the company to restore global access to Fable 5 and resume limited, vetted distribution of Mythos 5.

Regulatory Reversal and Timeline

Anthropic introduced Claude Fable 5 and Claude Mythos 5 on or about June 9 as part of its Claude AI platform. Fable 5 was positioned as the broadly accessible version of the Mythos class, with enhanced safety measures.

On June 12, the Commerce Department issued a private export-control directive requiring Anthropic to obtain U.S. permission before any foreign national—including non-citizen employees—could access Mythos 5 or Fable 5. Citing national security concerns, officials worried the models could be misused to exploit computer vulnerabilities. Because Anthropic lacked a reliable way to verify nationality in real time, it disabled access worldwide, effectively shutting down both models for roughly 18–19 days.

The export controls followed reports of a jailbreak technique against Fable 5’s cybersecurity safeguards, demonstrated by Amazon researchers and flagged to U.S. authorities. Officials feared Mythos 5 could identify zero-day vulnerabilities and build exploits, while Fable 5 might be manipulated to bypass guardrails.

Anthropic said the jailbreak revealed a small number of previously known, minor vulnerabilities rather than unique Mythos-level offensive capabilities. The company had not observed any compromise of Fable 5’s safeguards on cybersecurity or biology tasks.

On June 26, Commerce Secretary Howard Lutnick authorized Anthropic to provide Mythos 5 to a select group of vetted U.S. organizations, including companies and federal agencies that operate and defend critical infrastructure. Lutnick’s letter cited “appropriate safeguards” and conditions around proactive risk detection, collaboration on standards, and reporting of malicious activity.

On June 30, Commerce notified Anthropic that export controls were withdrawn and an export license was no longer required, enabling the company to begin restoring access.

Restoration, Distribution, and Safeguards

Fable 5 returned to users worldwide on July 1, reappearing on the Claude Platform, Claude.ai, Claude Code, and Claude Cowork interfaces. Anthropic said it will work to re-enable Fable 5 on Amazon Web Services, Google Cloud, and Microsoft Foundry marketplaces as soon as possible.

The company set a temporary allocation for paid customers: through July 7, Fable 5 may account for up to 50% of weekly usage limits on Claude Pro, Max, Team, and selected enterprise plans.

Mythos 5 access is being restored in a limited fashion—initially to roughly 100 U.S. companies and federal agencies via Anthropic’s Glasswing program. Anthropic said it will work with the U.S. government to expand access to additional domestic and international partners under ongoing safeguards.

Anthropic positions Mythos 5 as its most advanced model for businesses and cybersecurity professionals, with fewer guardrails and capabilities to identify vulnerabilities in code and build exploits. These higher-risk functions explain why access remains limited to vetted institutions.

Fable 5 serves as a public counterpart with heavy safety classifiers and extra guardrails. The model automatically falls back to an earlier version, Opus 4.8, on about 5% of sessions judged high risk.

The company retrained safety classifiers to target behaviors described in the jailbreak report. The federal Center for AI Standards and Innovation stress-tested the new safeguards, which have blocked the reported technique in more than 99% of cases.

Anthropic agreed to proactively detect and address security risks associated with its models, collaborate with U.S. agencies on standards for future releases, and inform authorities of malicious activity. These commitments contributed to Commerce’s decision to withdraw the export-control directive.

Enterprises and cybersecurity experts say the interruption and rapid restoration highlight how advanced AI models have become central to critical infrastructure planning. The episode underscored dependence on centralized model access and suggested Glasswing could serve as a controlled pathway for defensive testing with powerful AI.