Anthropic accuses Alibaba of stealing its AI with twenty-five thousand fake accounts

Published on June 28, 2026 | Translated from Spanish

Anthropic has accused Alibaba of creating 25,000 fake accounts to extract capabilities from its Claude model without paying. This massive distillation attack allows foreign companies to use American technology without investing, affecting local competition and the development of their own innovation.

Photorealistic technical illustration showing a massive server room with rows of glowing server racks, a hooded figure in the background controlling multiple screens displaying 25,000 small profile icons being created simultaneously, data streams flowing from a central AI core labeled Claude toward a distant Alibaba cloud logo, digital extraction tools crawling across neural network diagrams, red warning alerts flashing on monitors, dramatic blue and orange lighting, metallic surfaces reflecting code, cinematic security breach scene, ultra-detailed hardware components, glowing circuit traces, tense atmosphere.

AI Distillation: How Model Theft Works 🛡️

Distillation involves bombarding a model with queries to replicate its responses in one's own system. In this case, the fake accounts simulated legitimate users to extract knowledge from Claude. This method reduces development costs but violates terms of use and intellectual property rights. The Alibaba case sets a global precedent.

Chinese Competition Learns to Save at the AI Supermarket 🛒

It seems Alibaba decided that paying for technology is a thing of the past. With 25,000 accounts, they set up an AI supermarket where everything is free, you just need to know how to fill the cart. Now Anthropic is calling for stricter laws, while others look sideways to see if the deal repeats. Innovation has never been so cheap.