Anthropic has told US lawmakers in a June 10 letter that Alibaba's Qwen AI lab illicitly extracted Claude's capabilities using roughly 25,000 fake accounts, calling it the largest known distillation attack to date.
June 10, 2026 · Anthropic
"The largest known distillation attack on Anthropic to date"
Anthropic told US lawmakers and the White House that operators linked to Alibaba and its Qwen lab "brazenly" and "illicitly" extracted Claude's capabilities — using roughly 25,000 fraudulent accounts to harvest its most valuable coding and agentic-reasoning skills.
28.8M
exchanges generated with Claude
~25,000
fraudulent accounts used
45
days, Apr 22 – Jun 5, 2026
Scale of alleged extraction, by case
Exchanges with Claude (millions) — column height proportional to volume
Alibaba / QwenJun 2026
MiniMaxFeb 2026
Moonshot AIFeb 2026
DeepSeekFeb 2026
How an adversarial distillation attack works
Bypassing access controls to clone capabilities cheaper and faster than building them.
1 · Fake accounts
→
2 · Mass querying
→
3 · Train clone
Why it matters
First time Anthropic has publicly named a major Chinese tech conglomerate, not just a startup.
Risk of capability proliferation — potentially to military and intelligence uses — without original safeguards.
Comes days after Alibaba was added to the Pentagon's Chinese Military Companies list on June 8, 2026.
Anthropic urges antitrust clarity for info-sharing, penalties, export controls, and industry-government coordination.
Note: Alibaba did not respond to requests for comment. No broad confirmed reports of direct impact from the alleged campaign have surfaced, and many user observations remain unverified.
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article…