Trump Lifts Export Controls on Anthropic Claude

Anthropic's Claude Fable 5 and Mythos 5 models are back online after the U.S. government lifted export controls that had shut them down worldwide for 18 days. The reversal, announced Tuesday, restores access starting Wednesday and closes a standoff that briefly cut off every user of Anthropic's two flagship models, foreign and domestic alike.

Key Takeaways

• Fable 5 returns to Claude.ai, Claude Code, Claude Cowork and the Claude platform starting Wednesday.
• Pro, Max, Team and select Enterprise subscribers get Fable 5 access covering up to half their weekly usage allotment through July 7.
• Mythos 5 access resumed June 26 for a limited set of U.S. organizations; broader rollout through the Glasswing program is still under discussion with federal officials.
• Anthropic built a new safety classifier that blocks the flagged exploit technique in over 99 percent of test attempts.
• The episode landed as Anthropic pursues an IPO with a disclosed $47 billion revenue run rate and $965 billion valuation.

What triggered the shutdown

The Commerce Department's order, issued June 12, followed Amazon researchers documenting a method for coaxing Fable 5 into producing dangerous outputs tied to software vulnerability discovery. The rule barred access for foreign nationals broadly, a definition wide enough to sweep in Anthropic's own research staff. Rather than attempt a partial fix under that constraint, Anthropic concluded a full shutdown of both models was the only way to comply. The company noted that its internal testing found the same exploit reproducible on other widely deployed models, including OpenAI's GPT-5.5, suggesting the vulnerability wasn't unique to Fable 5's architecture.

The fix and the sign off

Anthropic's response was a new safety classifier layered on top of existing guardrails. In testing, it stopped the Amazon documented bypass in more than 99 percent of attempts. The Commerce Department's Center for AI Standards and Innovation reviewed both the original safeguards and the new classifier and confirmed the results, according to Anthropic. Any request that still trips the block gets rerouted to Claude Opus 4.8 rather than failing outright, keeping continuity for users mid workflow.

Commerce Secretary Howard Lutnick said on X that his department spent the two weeks prior working directly with Anthropic to align safety standards across government while preserving U.S. leadership in the field. White House chief of staff Susie Wiles echoed that framing, telling X that the shared goal is deploying capable models