The AI developer stated it has implemented technical safeguards over the last few months to reduce the success of these distillation efforts. Distillation occurs when a company uses a more advanced model to generate high-quality data to train its own smaller or separate model. Anthropic has published a technical post detailing the methods it now uses to block such attempts.
Wednesday, October 21, 2026
Story
Moonshot AI Secretly Routed Kimi Requests to Claude for Chinese Military Surveillance
AI8 days ago
Anthropic detected a series of distillation attacks where its Claude models were used as a hidden backend for a competitor's AI. In one instance, a user suspected of being affiliated with China's People's Liberation Army used Moonshot AI's Kimi models to track an individual across hundreds of CCTV cameras in Chengdu, not knowing the requests were being secretly forwarded to Claude, Anthropic reported. The diversion was part of Moonshot's effort to extract intelligence from Claude to improve its own models.
Looking for more context or analysis?
Tilt chat provides reliable, tailored insights, alerts, and automations.