Anthropic restores Fable 5 access, clamps down on Chinese AI use

The U.S. Commerce Department lifted an emergency export‑control order on Anthropic’s flagship Claude models Fable 5 and Mythos 5 on July 1, after a June ban was triggered by a security test that showed the model could generate exploit code. Anthropic demonstrated that the same capability existed in competing models and added a safety classifier that blocks the technique in over 99 % of cases.

At the same time, Anthropic announced a tightening of its access rules for Chinese entities. In a letter to U.S. senators, the company alleged that Alibaba’s Qwen AI lab conducted the largest known “distillation attack,” using about 25,000 fraudulent accounts to make roughly 28.8 million queries to Claude between April 22 and June 5, 2026. Anthropic is now enforcing identity verification, blocking payments from Chinese banks and shutting down proxy‑based workarounds used by firms such as ByteDance and Ant Financial.

Alibaba responded by banning the Claude Code tool for its staff, citing hidden steganographic code that flagged Chinese users. The ban, effective July 10, pushes the company to adopt its own Qoder coding assistant. The episode highlights growing U.S.–China tensions over advanced AI models and has prompted many enterprises to diversify their AI‑model strategies after the brief blackout of Fable 5.