AI spy war: Anthropic says rival labs used thousands of fake accounts to steal Claude's brain. Its new defences cut the theft by up to 99%, with tens of thousands of accounts banned.

Anthropic says its new defenses cut data harvested by distillation attacks by roughly 90 to 99 percent, using a two-stage classifier, preserved thinking that blocks new API accounts from editing prior context, and account bans. Anthropic reports banning tens of thousands of accounts, often within two hours, and says extracted reasoning tokens fell 10 to 100 times. It cites Zhipu's GLM-5.3 and Moonshot's Kimi, and acknowledges trade-offs: integration updates, a sub-0.1% false positive rate, and charging for blocked requests.

✨ AI summary of the linked source — not editorial content.

Explore the topic

• War
• Anthropic
• Claude
• AI