Quick Summary
- Fable 5 and Mythos 5 went offline on June 12 under a federal export control directive.
- The Commerce Department lifted the restrictions on June 30, ending an 18-day disruption.
- The order followed an Amazon-reported technique for bypassing one of Fable 5’s cybersecurity safeguards.
- Anthropic retrained its safety classifier and says the technique is now blocked in more than 99% of attempts.
- Mythos 5 was cleared first, on June 26, for a select group of trusted U.S. organizations; Fable 5 followed on June 30 for global rollout.
- Anthropic agreed to closer coordination with federal agencies on threat sharing and jailbreak evaluation standards going forward.
- The episode highlights that access to frontier AI models can now depend on regulatory decisions, not just technical readiness.
Fable 5 and Mythos 5, two of Anthropic’s most capable AI models, spent 18 days unavailable to users worldwide. The shutdown began June 12, 2026, after the U.S. Department of Commerce issued an export control directive tied to national security concerns. On June 30, the government reversed that order, and Anthropic began restoring access the following day.
The interruption affected far more than one company’s product line. Businesses that had built workflows around these models, including coding and customer-support automation, lost access without warning. The episode is a clear example of a risk enterprise AI buyers now have to plan around: regulatory decisions that can take a model offline overnight, independent of how well it performs technically.
What Happened to Fable 5 and Mythos 5
Fable 5 and Mythos 5 are both part of Anthropic’s Mythos-class model tier, launched June 9, 2026. Fable 5 is the public-facing, broadly available version, built with additional safety layers on top of the underlying model. Mythos 5 shares that same underlying architecture but carries fewer of those added restrictions, making it more capable in cybersecurity-related tasks — which is why it was limited to a smaller group of vetted organizations rather than released broadly.
Both models were suspended at the same time because of how the export order was written. It required Anthropic to block access for foreign nationals, whether inside or outside the United States, including the company’s own non-citizen employees. Because there was no reliable way to verify a user’s nationality in real time, Anthropic shut down access for everyone, everywhere, to comply.
Enterprise customers felt the effects immediately, and some organizations reportedly turned to alternative models, including lower-cost systems from Chinese developers, to bridge the gap while they waited.
Why the Government Pulled the Plug
The chain of events started with a security finding: researchers at Amazon reported a method for bypassing one of Fable 5’s cybersecurity safeguards. After the government became aware of the report, it issued the export control directive covering both models while the risk was reviewed.
Anthropic disputed the severity of the finding throughout the episode, arguing it didn’t warrant a full global shutdown. After reviewing the research with Amazon and federal officials, the company concluded the technique represented a boundary case in Fable 5’s safeguards rather than exposure of any Mythos-level capability unique to the more powerful model. Anthropic retrained its safety classifier in response, saying the reported technique is now blocked in more than 99% of attempts, while acknowledging the tighter filtering increases false positives on some legitimate coding requests.
How Anthropic Got the Models Back Online
The restoration happened in two stages. On June 26, Commerce Secretary Howard Lutnick authorized a limited release of Mythos 5 to a select group of trusted U.S. organizations, including government agencies and companies in critical infrastructure. Fable 5 remained offline at that point, since it serves a much larger public audience.
On June 30, Commerce lifted the export controls on both models entirely. Anthropic confirmed the reversal on X, thanking users for their patience during the restoration effort. Lutnick said the administration had worked closely with Anthropic over the prior two weeks to align on safeguards before approving Fable 5’s return. As part of the resolution, Anthropic agreed to closer coordination with federal agencies going forward, including proactive security-risk reporting and work toward shared industry standards for evaluating jailbreak techniques.
Fable 5 began rolling out globally on July 1 across the Claude platform, Claude.ai, Claude Code, and Claude Cowork, with access on AWS, Google Cloud, and Microsoft Foundry following shortly after. Usage is being capped at up to 50% of normal weekly limits for some plans through July 7 before returning to full availability.
What It Means If You’re Building on These Models
For organizations that rely on frontier AI models, this episode offers a few durable lessons.
A working model isn’t the same as an available one. Fable 5 and Mythos 5 never stopped functioning technically — they got taken offline by a decision made outside Anthropic entirely. That’s a different kind of risk than an outage or a bug, and it’s worth planning for separately.
Spreading across clouds didn’t help here. The restriction was on the model itself, not on AWS or Google Cloud or Azure. Teams running Fable 5 through three different providers still lost all three at once.
Being a “trusted partner” already paid off. Organizations that had already gone through government vetting got Mythos 5 back four days before everyone else got Fable 5. If that kind of access matters to your business, it’s worth applying for before you need it, not after.
Final Thoughts
For most users, this is already over. Fable 5 and Mythos 5 are back, and day-to-day work should feel normal again within a few days. But the underlying thing that just happened doesn’t undo itself: a US agency can now take a frontier model offline worldwide with about a day’s notice, and that’s true regardless of how the model performs or how many people rely on it.
That’s not a reason to panic about using Claude or any other frontier model. It is a reason to treat “who can pull access, and how fast” as a real question when you’re deciding what to build your workflows around, right alongside cost and capability.
Discover how AI is reshaping technology, business, and healthcare—without the hype.
Visit InfluenceOfAI.com for easy-to-understand insights, expert analysis, and real-world applications of artificial intelligence. From the latest tools to emerging trends, we help you navigate the AI landscape with clarity and confidence
