Dave Jimenez
Anthropic, the AI research company behind the Claude family of models, builds large language systems for consumer chat, coding and enterprise deployment. The company said Tuesday that Washington has lifted export controls on its Claude Fable 5 and Mythos 5 models, ending an 18 day shutdown that had cut off every user worldwide from both systems.
Key Takeaways
- Fable 5 returns Wednesday across Claude.ai, Claude Code, Claude Cowork and the Claude platform after an 18 day export ban.
- Anthropic built a new safety classifier that reportedly stops the flagged exploit technique in over 99 percent of attempts.
- Reinstatement on AWS, Google Cloud and Microsoft Foundry is planned but has no confirmed date.
- Mythos 5 access resumes only for a limited set of U.S. organizations following a June 26 government approval.
- The disruption landed as Anthropic pursues a confidential IPO filing that discloses a $47 billion revenue run rate and a $965 billion valuation.
Why Washington Pulled Fable 5 Offline
The original order came down June 12 after Amazon researchers documented a way to coax dangerous outputs from Fable 5 tied to software vulnerability discovery. Because the restriction applied to all foreign nationals, a group that included researchers inside Anthropic itself, the company decided full shutdown was the only way to stay compliant. Anthropic has since noted that its internal testing turned up the same exploit path in other widely used systems, including OpenAI's GPT-5.5, suggesting the vulnerability was not unique to its own architecture.
The Fix and the Government Review
Anthropic responded by engineering a new safety classifier that, according to its own testing, blocks the Amazon documented bypass in more than 99 percent of attempts. The Commerce Department's Center for AI Standards and Innovation reviewed both the pre existing guardrails and the added classifier and confirmed they held up. Until Fable 5 access is broadly restored on cloud partners, blocked requests will route instead to Claude Opus 4.8, keeping continuity for customers who rely on the model line for production workloads.
Restoring Access: Timeline and Gaps
Wednesday's reinstatement covers Claude.ai, Claude Code, Claude Cowork and the wider Claude platform. Through July 7, subscribers on Pro, Max, Team and select Enterprise tiers get Fable 5 access worth up to half their weekly usage allotment before credits kick in. Cloud availability through AWS, Google Cloud and Microsoft Foundry is coming, Anthropic says, but no date has been set. Mythos 5 access is narrower still: only a limited group of domestic organizations regained entry after the June 26 approval, and expanding that to more partners at home and abroad through the Glasswing program remains under discussion with federal officials.
Political Signals and Corporate Stakes
Commerce Secretary Howard Lutnick said on X that his department spent two weeks vetting Fable 5 with Anthropic to align the government's position and reinforce American standing in AI. White House chief of staff Susie Wiles echoed that framing, saying the shared goal is deploying strong technology quickly and safely, according to reporting from The Wall Street Journal. Notably, Lutnick's correspondence during the review was addressed to co-founder Tom Brown rather than CEO Dario Amodei, and CNBC has reported Brown took the lead negotiating role. That leadership shift comes as Anthropic has filed a confidential IPO prospectus showing a $47 billion revenue run rate and a $965 billion valuation, a backdrop that made the 18 day outage especially costly in timing.
What Happens to Mythos 5 Access Next
The bigger unresolved question is how far Mythos 5's clearance will extend beyond the current limited group of U.S. organizations. Whether the Glasswing program eventually opens access to additional domestic and international partners depends on continuing talks with federal authorities, and Anthropic has given no timeline for that outcome or for full multicloud restoration of Fable 5.
Carlos Dixon
