Anthropic Accuses Alibaba of Using Claude Outputs to Train Its Models While Revoking Access to Its New Model Over U.S. Export Rules

Anthropic accused Alibaba of using outputs from its Claude model to train rival systems through distillation in the week of June 26, 2026. The company has previously stated that other Chinese firms have employed the same technique. Anthropic released Fable 5, a safeguarded version of its Mythos model, in early June 2026.

It revoked worldwide access to the model a few days later in response to export controls imposed by the Trump administration, Wired reported. The company rolled out identity verification for some Claude users in April 2026. The process, handled by Persona, requires upload of a government-issued photo ID such as a passport or driver’s license, and IDs from unsupported countries do not satisfy the check.

Anthropic does not offer commercial access to Claude in China or to subsidiaries of Chinese companies located outside China. Dario Amodei, Anthropic’s cofounder and CEO, has singled out Chinese access to frontier models as a critical threat to U.S. national security.

Michael Aciman, an Anthropic spokesperson, stated that the company uses evolving detection systems including identity verification to enforce policies against unauthorized access. He added that Anthropic has worked to detect and disrupt proxy networks providing Claude access in China.

Chinese users access the model through transfer stations that purchase API access outside China and redistribute tokens inside the country.

These services often charge less than Anthropic’s direct prices because they obtain enterprise discounts. Justin Sun opened his own transfer station in May 2026. Zilan Qian, a research associate at the Oxford China Policy Lab, stated that Chinese software developers overwhelmingly prefer Claude Code and OpenAI’s Codex over domestic tools.