U.S. Government Lifts Export Controls on Claude Fable 5 and Mythos 5, Ending 18-Day Standoff
The intense regulatory standoff between Anthropic and the Trump administration has concluded. On June 30, 2026, the U.S. Department of Commerce officially withdrew its emergency export controls on Anthropic's flagship models, Claude Fable 5 and Mythos 5, ending an 18-day freeze that had forced the models offline12 due to severe cybersecurity concerns.34
Following the lifting of these restrictions, Anthropic officially redeployed Claude Fable 5 globally on July 1, 2026. However, the model did not return unchanged. To satisfy government regulators and prevent misuse, Anthropic implemented a robust new real-time "cybersecurity safety classifier" with a highly conservative "safety margin." This classifier is trained to detect, monitor, and block dual-use cybersecurity activities, erring on the side of caution by blocking benign requests if they fall within the safety margin.5
Additionally, Anthropic and its Glasswing partners proposed a standardized "Cyber Jailbreak Severity (CJS)" scale (ranging from CJS-0 to CJS-4) to systematically grade the severity of AI jailbreaks based on capability gain, breadth, ease of weaponization, and discoverability. To crowd-source defenses, Anthropic also launched a specialized HackerOne bug bounty program specifically targeting cyber jailbreaks in Fable 5.
"Claude Fable 5 has been re-deployed and is now available globally for all users... These are the AI systems that accompany the model that detect and block dangerous (or potentially dangerous) cybersecurity uses." — Anthropic Newsroom
"In the meantime, we’re also launching a new HackerOne program where security researchers can submit potential cyber jailbreaks they’ve discovered in Fable 5..." — Anthropic Newsroom
This successful resolution allows Anthropic to continue scaling its enterprise agent offerings, such as the newly released Anthropic Launches Claude Sonnet 5 and Claude Science to Drive Enterprise Agent Efficiency, while establishing a precedent for how frontier labs coordinate real-time safety patches and risk metrics with federal oversight.
-
An instance of National security mandates are turning public frontier AI launches into gated, state-vetted handovers. — The federal government lifted its emergency export controls on Anthropic only after the models were vetted and restricted with defensive classifiers. ↩︎
-
An instance of National security now mandates the forced localization of AI development. — This event demonstrates how national security agencies are actively using emergency export controls to regulate and vet the rollouts of advanced frontier AI models. ↩︎
-
An instance of National security oversight is converting software releases into high-friction diplomatic negotiations. — The direct activation and suspension of commercial frontier models by the Department of Commerce demonstrates that software capability releases have become high-friction geopolitical events. ↩︎
-
An instance of Geopolitical sovereignty now dictates the release of frontier AI models. — Federal agencies are directly treating software models as national security assets, freezing and releasing international distribution under geopolitical pretenses. ↩︎
-
An instance of Securing AI systems requires sacrificing agentic autonomy and zero-data-retention privacy. — To secure federal clearance and block dangerous use cases, the model enforces highly restrictive real-time safeguards that preemptively block even benign, normal requests. ↩︎