Anthropic says three Chinese labs trained rivals on its model outputs

single source · 1 articles · safety · confidence: low · first seen 2026-09-10 20:57 UTC

Anthropic published a report on Thursday, 10 September 2026, alleging that Alibaba, Moonshot AI, and DeepSeek have run persistent distillation attacks against its models. Distillation means training one model on another model’s outputs. The report says these efforts have escalated in recent months as competition has intensified. It does not provide evidence beyond the claim, and the named companies have not publicly responded. The report is Anthropic’s own account; no independent verification has been published.

What this means for you

Nothing to act on yet. This is a single company’s allegation, not a legal finding or a technical disclosure. Watch for responses from the named labs and any policy or licensing changes from Anthropic.

Key facts

  • ·Anthropic released a report on Thursday, 10 September 2026, alleging persistent distillation attacks by China-based AI companies. source
  • ·The companies named are Alibaba, Moonshot AI, and DeepSeek. source
  • ·Anthropic says the attacks have escalated in recent months as competition intensified. source

What the sources say

  • TechCrunch AIReports Anthropic's allegation that three Chinese labs ran distillation campaigns against its models.

Sources

The original reporting. Follow these — they did the work.

← the wire