Anthropic Distillation Defenses Create New SaaS Security Category — Behavioral Fingerprinting for APIs

Anthropic·medium signal

Anthropic revealed industrial-scale distillation attacks — 24K+ fraudulent accounts, 16M+ exchanges from DeepSeek, Moonshot AI, and MiniMax extracting Claude's capabilities. Attacks look like normal usage (no traditional hack required), making detection hard. Anthropic built behavioral fingerprinting classifiers that detect chain-of-thought elicitation patterns and coordinated activity. The response: tighter rate limiting, anomaly detection, ToS enforcement across the industry. **Builder opportu