Anthropic Says It Detected Industrial-Scale Claude Distillation Campaigns
Anthropic said three labs used thousands of accounts and millions of exchanges in attempts to extract Claude capabilities.
Anthropic said it had identified campaigns by DeepSeek, Moonshot and MiniMax that attempted to extract Claude capabilities through large volumes of interactions. The company described more than 16 million exchanges across roughly 24,000 fraudulent accounts.
The report sharpened a practical issue for frontier labs: API abuse can become a model-development shortcut for rivals. Anthropic distinguished legitimate self-distillation from unauthorized extraction, arguing that access controls, fraud detection and usage monitoring were now part of protecting model capability.
Editorial sources
Every claim in this briefing traces back to the references below.
- Detecting and preventing distillation attacks — Anthropic — Primary company report https://www.anthropic.com/news/detecting-and-preventing-distillation-attacks
