Surveillance:
Cybersecurity Books

Threats, defenses, and the human habit of clicking things we really shouldn't.

Evaluate

Sponsorship confirmed.

Anthropic Accuses Alibaba of Major AI Model Theft, Urges Export Controls

Jun 25, 2026
Anthropic Accuses Alibaba of Major AI Model Theft, Urges Export Controls

Anthropic has made serious allegations against Alibaba, claiming that the Chinese tech giant executed the largest known distillation attack on its AI models. In a letter dated June 10 to Senators Tim Scott and Elizabeth Warren, Anthropic detailed a systematic effort by Alibaba’s AI lab to siphon intelligence from its models, involving approximately 25,000 fraudulent accounts and resulting in 28.8 million exchanges between April 22 and June 5.

Distillation attacks involve training a smaller, less expensive AI model using the outputs of a more powerful one, effectively replicating its capabilities without the associated research and development costs. For Anthropic, this issue transcends intellectual property; it poses a significant national security threat. The company argues that such distillation campaigns directly undermine U.S. restrictions on advanced AI chips. If Chinese firms can extract knowledge from American models, the effectiveness of export controls on the hardware used to train those models diminishes.

Surveillance:
Cybersecurity Books
Evaluate
Geopolitical Maps & Reference Atlases
Evaluate
Intelligence History
Evaluate

Sponsorship confirmed.

This is not the first time Anthropic has raised concerns about distillation attacks. Earlier this year, the company reported similar campaigns by various Chinese AI labs, including DeepSeek, Moonshot, and MiniMax, which collectively involved over 16 million exchanges through around 24,000 fake accounts.

Anthropic's letter coincided with a broader context of U.S. government action. A White House memorandum issued in April expressed support for initiatives aimed at detecting and mitigating distillation attacks. Anthropic noted that Alibaba had previously ignored warnings from the Trump Administration regarding such practices.

Following the letter, on June 12, the U.S. Commerce Department released an export control directive that led Anthropic to suspend global access to its advanced AI models, Fable 5 and Mythos 5. The company is now advocating for coordinated efforts among industry players and cloud providers to combat these illicit distillation activities, emphasizing the need for a united front against such threats to national security.

Surveillance:
Cybersecurity Books

Threats, defenses, and the human habit of clicking things we really shouldn't.

Evaluate
Geopolitical Maps & Reference Atlases

Political maps and reference atlases for keeping the geography straight. Check the edition before treating a border as current.

Evaluate

Sponsorship confirmed.

Recommended for You
OpenAI and Anthropic Advocate for Binding AI Regulations in the UK
Intelligence Oct 13, 2026

OpenAI and Anthropic Advocate for Binding AI Regulations in the UK

OpenAI and Anthropic have urged UK lawmakers to implement binding regulations for powerful AI systems, emphasizing the need for mandatory…

Surveillance:
Cybersecurity Books

Threats, defenses, and the human habit of clicking things we really shouldn't.

Evaluate
Geopolitical Maps & Reference Atlases

Political maps and reference atlases for keeping the geography straight. Check the edition before treating a border as current.

Evaluate
Intelligence History

Secret operations, double agents, and the decisions left out of the official version.

Evaluate
World Affairs Books

Diplomacy, rivalries, and the backstory behind the next international headline.

Evaluate
Intelligence Analysis & Critical Thinking

Books on evaluating evidence, testing assumptions, and building sound assessments. Confidence is not a substitute for a source.

Evaluate
Counterintelligence Case Studies

Nonfiction accounts of moles, deception, and investigations. The warning signs tend to look clearer in the final report.

Evaluate

Sponsorship confirmed.