Krux
February 24, 2026
Anthropic Blocked 16 Million Illicit Queries from Chinese AI Labs
Published: February 24, 2026 at 1:05 AM
Updated: February 24, 2026 at 1:05 AM
100-word summary
Anthropic disclosed it detected and shut down three industrial-scale campaigns by DeepSeek, Moonshot AI, and MiniMax to illicitly clone Claude's capabilities through 16 million exchanges across 24,000 fraudulent accounts. The labs used sprawling proxy networks to extract reasoning patterns, coding skills, and tool use via coordinated queries that shifted within 24 hours when new Claude models dropped. Anthropic warns distillation attacks circumvent export controls and strip safety guardrails, creating risks for cyber operations and disinformation. The company is urging industry-wide intelligence sharing and deploying behavioral fingerprinting to catch coordinated account activity before capabilities leak to foreign competitors.
What happened
Anthropic disclosed it detected and shut down three industrial-scale campaigns by DeepSeek, Moonshot AI, and MiniMax to illicitly clone Claude's capabilities through 16 million exchanges across 24,000 fraudulent accounts. The labs used sprawling proxy networks to extract reasoning patterns, coding skills, and tool use via coordinated queries that shifted within 24 hours when new Claude models dropped. Anthropic warns distillation attacks circumvent export controls and strip safety guardrails, creating risks for cyber operations and disinformation.
Why it matters
The company is urging industry-wide intelligence sharing and deploying behavioral fingerprinting to catch coordinated account activity before capabilities leak to foreign competitors.