Anthropic Accuses Alibaba of Distillation Attack

Anthropic accused Alibaba in a June 10, 2026 letter of running a large-scale distillation campaign through Alibaba’s Qwen AI lab, a move that could increase U.S. export-control scrutiny and heighten regulatory risks for technology investors.

Largest Known Distillation Campaign on Claude AI

Anthropic, the U.S. developer of the Claude AI models, said operators affiliated with Alibaba Group Holding Ltd. and its Qwen AI lab conducted what it described as the largest known distillation attack on Claude. The campaign ran from April 22 to June 5, 2026, involving nearly 25,000 fraudulent accounts and about 28.8 million interactions with Claude models. The operation targeted Claude’s most advanced capabilities, including software engineering and agentic reasoning, which involves complex task planning and autonomous agent behavior.

Anthropic defines distillation as training a less capable model using outputs from a more powerful system. It classifies such campaigns as adversarial or illicit when they rely on fake accounts or violate terms of service. The company warned that adversarial distillation enables rivals to replicate leading U.S. AI models at a fraction of the training and research cost, often producing systems without original safety guardrails. This process could accelerate competitors’ access to Anthropic’s advanced Mythos Preview capabilities.

In February, Anthropic disclosed earlier industrial-scale distillation efforts by Chinese AI startups DeepSeek, Moonshot AI, and MiniMax. Those campaigns logged over 150,000 exchanges for DeepSeek, more than 3.4 million for Moonshot AI, and over 13 million for MiniMax. Anthropic said these operations were growing in intensity and sophistication, calling for rapid, coordinated action among industry and policymakers.

Policy and Regulatory Developments

Anthropic sent the June letter to Senators Tim Scott (R-S.C.), chair of the Senate Banking Committee, and Elizabeth Warren (D-Mass.), the committee’s ranking member. The letter was also circulated to White House officials and other senators ahead of a Senate Banking Committee hearing on artificial intelligence.

In April 2026, the White House Office of Science and Technology Policy issued a memo committing U.S. agencies to assist private AI labs in identifying and coordinating against large-scale distillation attempts, including threat-intelligence sharing and joint exercises. On June 12, 2026, the U.S. Commerce Department imposed export controls on Anthropic’s Mythos and Fable frontier models, citing national security concerns that they could be used by military intelligence in China and other countries of concern. Following the order, Anthropic disabled global access to the affected models.

Anthropic urged the government to clarify antitrust guidelines so U.S. labs can share information about distillation attempts without violating competition laws. It also called for penalties, including sanctions or blacklisting, against firms that use illicit distillation and recommended maintaining or strengthening export controls on advanced AI chips as part of broader defensive measures. Lawmakers are considering related steps: Senators Bill Hagerty (R-Tenn.) and Andy Kim (D-N.J.) are reportedly drafting an amendment to defense legislation to blacklist or sanction firms improperly exploiting U.S. model outputs. A bipartisan House measure backed by Representatives Bill Huizenga and Sydney Kamlager-Dove is also under consideration. U.S. officials estimate unauthorized distillation costs Silicon Valley labs billions of dollars, framing it as a significant strategic and economic risk.

Anthropic said it will continue working with Congress and the administration to address illicit distillation while pursuing commercial deployment under evolving export-control constraints.