Anthropic Urges Senate to Curb Alibaba AI Copying via 25,000 Accounts
Updated
Updated · The New York Times · Jul 6
Anthropic Urges Senate to Curb Alibaba AI Copying via 25,000 Accounts
3 articles · Updated · The New York Times · Jul 6
Summary
A June 10 letter from Anthropic asked Senators Tim Scott and Elizabeth Warren to examine ways to curb Chinese A.I. distillation after accusing Alibaba of illicitly copying its technology.
Anthropic said Alibaba accessed its systems through tens of thousands of unauthorized accounts—25,000 in an earlier account—to harvest outputs and train its own models at industrial scale.
The complaint lands as U.S. companies warn China trails American A.I. by only about six months and argue distillation is narrowing that gap in areas from drug research to military uses.
Z.ai’s new GLM-5.2 model has added urgency by approaching top U.S. systems and rivaling them in cybersecurity, a field Washington and American labs view as strategically critical.
When AI models can be 'distilled,' is it intellectual property theft or just aggressive market competition?
Is the U.S. creating a more formidable AI rival by trying to cut China off from its technology?
Anthropic vs. Alibaba: Inside the 16 Million-Exchange AI Distillation Attack and Its Global Fallout
Overview
In June 2026, Anthropic accused Alibaba of running an 'industrial-scale' AI distillation campaign to extract advanced capabilities from its Claude AI models, despite US government warnings about foreign copying of frontier technologies. This alleged campaign followed similar large-scale efforts previously identified from other AI labs, showing a growing trend of sophisticated attacks targeting leading AI models. Anthropic highlighted Alibaba's US ties and noted that the fraudulent accounts were linked to Alibaba's Qwen lab. The incident has intensified US-China technology tensions and prompted calls for stronger industry and government collaboration to defend against such threats.