Anthropic accused Alibaba of 28.8 million AI model extraction attempts
Why it matters: Anthropic told the Senate the activity used about 25,000 fraudulent accounts and could sharpen Washington's push for tighter AI controls.
Anthropic told the U.S. Senate Banking Committee that Alibaba and its AI lab tried to pull capabilities from Anthropic's systems through what it called the largest known distillation attack it has faced. In a June 10 letter to Sens. Tim Scott and Elizabeth Warren, the company said operators linked to Alibaba made 28.8 million exchanges with Anthropic models using roughly 25,000 fraudulent accounts between April 22 and June 5. Distillation is a training technique that uses outputs from a more powerful model to build a smaller one. Alibaba did not immediately respond to CNBC's request for comment. The complaint comes as Washington weighs tougher measures to help AI firms detect and coordinate against large-scale model extraction.