Cyber & Technology — 2026-06-26

Anthropic Accuses Alibaba of Largest Known AI Model Extraction Campaign Against Claude

BLUFAnthropic's Senate Banking channel converts a commercial distillation dispute into legislative record, pressuring the administration to act on PRC-linked AI extraction or own the inaction.

Anthropic's June 10 letter to Senate Banking Committee chairs Tim Scott and Elizabeth Warren, confirmed by CNBC, accused operators affiliated with Alibaba and its Qwen AI lab of conducting 28.8 million exchanges with Claude through approximately 25,000 fraudulent accounts between April 22 and June 5 12. The letter described the campaign as targeting Anthropic's Mythos Preview model's coding and digital security capabilities, characterizing it as the largest known distillation attack on the company to date 12. Alibaba had not responded to comment requests from CNBC or Global Times as of their respective publication times 13. Global Times cited Tian Feng, former dean of SenseTime's Intelligence Industry Research Institute, calling the allegations rooted in "technological hegemony anxiety" aimed at curbing China's AI development rather than addressing substantive IP concerns 3.

Analysis
The campaign's 28.8 million exchanges, nearly twice the combined extraction volume attributed to DeepSeek, Moonshot, and MiniMax in February, signal accelerating tempo against Anthropic's most sensitive systems. By routing the accusation through Senate Banking Committee chairs, Anthropic converts what would otherwise be a terms-of-service dispute into a legislative record the administration cannot ignore without cost. The "operators affiliated with" framing, however, leaves open the possibility that rogue API resellers conducted the campaign without Alibaba's corporate sanction, undercutting the espionage characterization. Attribution confidence is moderate, resting on behavioral and metadata analysis corroborated by the OSTP April memorandum but not independently verified. The underlying disclosure is a single interested-party letter, with Global Times providing adversarial counter-framing rather than independent technical corroboration.

4 sources