Anthropic launches Claude Sonnet 5 and brings back flagship models Fable and Mythos after an 18-day pause

Anthropic has ended its forced pause: on June 30, 2026, the company launched Claude Sonnet 5 and restored access to its flagship Fable and Mythos models, which had been shut down on June 12 at the request of US federal regulators as part of an export control review.

What happened on June 12

On June 12, 2026, US authorities issued an export control directive that directly affected Anthropic. The company had to temporarily suspend its most powerful systems. Fable and Mythos fell under the restrictions. Anthropic classes them as "frontier models": cutting-edge models that set the upper bound of performance in its product line. All other versions of Claude kept running normally during this period.

After 18 days of federal review, Anthropic received permission to resume operations. Claude Sonnet 5 officially launched alongside the restoration of access to the flagships.

Key facts about the incident:

• June 12, 2026: the federal export control directive takes effect
• 18 days: the length of the forced pause for Fable and Mythos
• June 30, 2026: the review ends and access to the flagship models is restored
• Claude Sonnet 5: a new model launched at the same time the flagships came back
• Fable and Mythos: Anthropic's frontier models, which form the top tier of the Claude lineup

Why regulators can halt AI models

US export control law increasingly extends to high-performance AI systems, especially those that could be used for strategic or military purposes. The specific grounds for the June 12 directive were not disclosed publicly: Anthropic limited itself to confirming the pause and announcing its end once the review was completed.

Under the general logic of export controls, the higher an AI system's performance, the more likely it is to fall into a regulated category. That explains why Fable and Mythos were chosen rather than the lighter models in the lineup. Export restrictions on AI chips introduced in previous years are now being complemented by similar oversight mechanisms for AI software, and Anthropic's 18-day pause showed clearly that this control already works in real time.

What Claude Sonnet 5 is and where it sits in the lineup

Sonnet 5 is a new entry in the Claude family. Anthropic traditionally positions Sonnet versions as a balance between performance and per-call cost: they are more powerful than the compact Haiku models but cheaper to run than the flagship Opus, Fable or Mythos. Developers and enterprise customers most often use Sonnet versions for tasks that need high quality without the most powerful and expensive models. Releasing Sonnet 5 together with the return of the flagships means the entire Claude product line is running normally again, after eighteen days of forced downtime.

What this means

The restoration of Fable and Mythos and the launch of Sonnet 5 return Anthropic to full operating capacity after an unusual regulatory episode. The precedent matters more, though: the 18-day pause showed that high-performance language models now sit within the scope of export controls alongside hardware chips. For enterprise customers who depend on flagships at the level of Fable and Mythos, this is a new class of operational risk: regulatory, and fundamentally outside the control of the developer itself.