Yes

>>Hírek

>>Stunning AI Spy Scandal: Anthropic Reveals 24,000 Fraudulent Accounts in Chinese Labs Exploiting Claude Amid Ongoing Chip Export Dispute

Stunning AI Spy Scandal: Anthropic Reveals 24,000 Fraudulent Accounts in Chinese Labs Exploiting Claude Amid Ongoing Chip Export Dispute

crypthub

about 3 hours ago

Distillation Attack Uncovered

Anthropic’s security team uncovered a coordinated campaign by three Chinese AI labs—DeepSeek, Moonshot AI, and MiniMax—using over 24,000 fake accounts to query Claude. The operation produced more than 16 million exchanges, focusing on Claude’s advanced reasoning, tool use, and coding capabilities. It unfolded as the United States debated loosening export controls on AI chips to China.

Ethical and Security Implications

Model distillation is legitimate for internal optimization, but applying it to a competitor’s proprietary system amounts to intellectual‑property theft. Illicitly distilled models lack the safety safeguards built into U.S. AI services, raising the risk of malicious uses such as cyber attacks or disinformation. Experts warn that this practice amplifies national‑security concerns and calls for clearer legal protections.

Calls for Coordinated Response

Anthropic urges a multi‑layered strategy that combines technical detection tools, industry standards, and stricter export controls on advanced chips. Policymakers are asked to limit chip sales to firms implicated in distillation attacks. Broad cooperation among AI companies, cloud providers, and governments is deemed essential to protect frontier models.