Anthropic00:00PolicyOfficial Blog
Anthropic details defenses vs distillation attacks
Understand practical controls to reduce AI model extraction abuse.
Key Points
- 1Explains illicit distillation/extraction attacks
- 2Detected ~24k fraudulent accounts
- 3Observed 16M+ exchanges
- 4Strengthens detection, sharing, access controls
Anthropic published its response to illicit “distillation” campaigns that attempted to extract Claude outputs to train competing models. It reports detecting roughly 24,000 fraudulent accounts and over 16 million exchanges. The post underscores that misuse defense is now a core part of product reliability and safety.