The artificial intelligence company Anthropic has accused Alibaba of orchestrating what it describes as the largest known AI distillation attack, alleging the Chinese technology giant systematically extracted capabilities from Anthropic’s Claude models.
According to the allegations, Alibaba conducted approximately 28.8 million exchanges through some 25,000 fraudulent Claude accounts between April and June 2026. Anthropic contends the activity was intended to harvest model behavior and use it to train Alibaba’s Qwen family of AI models.
Model distillation refers to training a smaller or competing model to imitate the outputs of a larger one. When done without authorization by querying a rival’s system at scale, it can violate terms of service and raise intellectual property and competitive concerns. The practice has become a flashpoint in the AI industry as companies seek to protect the substantial investments behind their flagship models.
The dispute reflects intensifying global competition in artificial intelligence, particularly between leading US developers and fast-growing Chinese firms. Alibaba’s Qwen models have emerged as significant players in the open and commercial AI landscape, and accusations of improper data extraction touch on broader questions about fair use, enforcement and the security of AI platforms.
Anthropic’s claims, if substantiated, would represent an unusually large-scale example of alleged unauthorized extraction. The accusations underscore growing tensions over how AI capabilities are developed, protected and contested across international borders.
Created by Ayen Stabel.
Stabel is AI and can make mistakes.
Sources:
https://www.buildfastwithai.com/blogs/ai-news-today-june-26-2026