Blab AI
Anthropic Accuses Alibaba of Illicitly Extracting Claude Capabilities Through Large-Scale Distillation
On 24 June 2026, Anthropic— the San Francisco‑based developer of the Claude series— sent a formal letter to U.S. senators and White House officials. The letter accuses Alibaba Group’s Qwen AI lab of mounting a large‑scale distillation campaign against Claude.
According to the document, Alibaba‑linked operators created almost 25,000 fake accounts and used them to exchange 28.8 million prompts with Claude from April to June 2026. The exchanges were designed to harvest Claude’s advanced functions, such as software‑engineering assistance and agentic reasoning. Anthropic called the effort “the largest known attempt by a Chinese company to piggyback on U.S. frontier AI models.”
Alibaba has not yet replied. An Anthropic spokesperson, who declined to elaborate on the letter, said the company seeks coordinated action between government and industry to counter what it labels “illicit, industrial‑scale distillation.”
The complaint arrives amid a broader U.S. AI‑industry alarm. In April, the White House Office of Science and Technology Policy warned that large‑scale distillation—using model outputs to train new systems—poses a national‑security threat. The memo differentiated the practice from legitimate research by noting its reliance on thousands of proxy accounts.
Anthropic noted that the Alibaba campaign began after the memo’s release and proceeded in defiance of the administration’s warning. The company cautioned that ignoring the issue could give China an AI advantage.
Legislative activity is mounting. Senators Bill Hagerty (R‑TN) and Andy Kim (D‑NJ) are reportedly drafting an amendment to a defense bill that would blacklist or sanction any Chinese firm that improperly accesses U.S. AI model output. A bipartisan House bill, supported by Representatives Bill Huizenga (R‑MI) and Sydney Kamlager‑Dove (D‑NJ), is also slated for consideration.
Earlier this month, the U.S. Defense Department added Alibaba to its blacklist over alleged ties to China’s armed forces. Alibaba has sued the Pentagon to have the designation removed.
Anthropic’s letter also references its own struggles with U.S. regulators. In January 2026, the Department of Defense barred all U.S. private‑military contractors from doing business with Anthropic after the firm refused to drop contractual prohibitions on mass domestic surveillance and autonomous weapons. The Commerce Department later imposed export controls that barred foreign persons from using Anthropic’s top models.
Valued at $965 billion in May 2026, Anthropic is preparing for an initial public offering. It has joined forces with OpenAI and Google in the Frontier Model Forum, a coalition that shares intelligence on distillation attacks by Chinese firms such as DeepSeek, Minimax, and Moonshot AI.
U.S. officials estimate that unauthorized distillation costs Silicon Valley labs billions of dollars. Anthropic is urging the government to clarify antitrust guidelines and to enable greater information sharing among U.S. companies to counter distillation.
The letter highlights the growing tension between U.S. AI developers and Chinese competitors. While Alibaba has not yet responded, Anthropic’s appeal to Washington signals a push for tighter export controls and regulatory oversight of AI model use.
The situation remains fluid. Anthropic’s calls for additional government support may encounter resistance from the White House, which has recently imposed export controls on the company’s top models. The outcome of pending defense legislation and the response to the letter will shape the regulatory landscape for AI development and cross‑border technology transfer.
According to the document, Alibaba‑linked operators created almost 25,000 fake accounts and used them to exchange 28.8 million prompts with Claude from April to June 2026. The exchanges were designed to harvest Claude’s advanced functions, such as software‑engineering assistance and agentic reasoning. Anthropic called the effort “the largest known attempt by a Chinese company to piggyback on U.S. frontier AI models.”
Alibaba has not yet replied. An Anthropic spokesperson, who declined to elaborate on the letter, said the company seeks coordinated action between government and industry to counter what it labels “illicit, industrial‑scale distillation.”
The complaint arrives amid a broader U.S. AI‑industry alarm. In April, the White House Office of Science and Technology Policy warned that large‑scale distillation—using model outputs to train new systems—poses a national‑security threat. The memo differentiated the practice from legitimate research by noting its reliance on thousands of proxy accounts.
Anthropic noted that the Alibaba campaign began after the memo’s release and proceeded in defiance of the administration’s warning. The company cautioned that ignoring the issue could give China an AI advantage.
Legislative activity is mounting. Senators Bill Hagerty (R‑TN) and Andy Kim (D‑NJ) are reportedly drafting an amendment to a defense bill that would blacklist or sanction any Chinese firm that improperly accesses U.S. AI model output. A bipartisan House bill, supported by Representatives Bill Huizenga (R‑MI) and Sydney Kamlager‑Dove (D‑NJ), is also slated for consideration.
Earlier this month, the U.S. Defense Department added Alibaba to its blacklist over alleged ties to China’s armed forces. Alibaba has sued the Pentagon to have the designation removed.
Anthropic’s letter also references its own struggles with U.S. regulators. In January 2026, the Department of Defense barred all U.S. private‑military contractors from doing business with Anthropic after the firm refused to drop contractual prohibitions on mass domestic surveillance and autonomous weapons. The Commerce Department later imposed export controls that barred foreign persons from using Anthropic’s top models.
Valued at $965 billion in May 2026, Anthropic is preparing for an initial public offering. It has joined forces with OpenAI and Google in the Frontier Model Forum, a coalition that shares intelligence on distillation attacks by Chinese firms such as DeepSeek, Minimax, and Moonshot AI.
U.S. officials estimate that unauthorized distillation costs Silicon Valley labs billions of dollars. Anthropic is urging the government to clarify antitrust guidelines and to enable greater information sharing among U.S. companies to counter distillation.
The letter highlights the growing tension between U.S. AI developers and Chinese competitors. While Alibaba has not yet responded, Anthropic’s appeal to Washington signals a push for tighter export controls and regulatory oversight of AI model use.
The situation remains fluid. Anthropic’s calls for additional government support may encounter resistance from the White House, which has recently imposed export controls on the company’s top models. The outcome of pending defense legislation and the response to the letter will shape the regulatory landscape for AI development and cross‑border technology transfer.
