Anthropic alleges 28.8M-exchange distillation-attack campaign against Claude Mythos Preview — April 22 to June 5 2026, attributed to prior efforts by DeepSeek, Moonshot AI, and MiniMax
Anthropic alleges a campaign of 28.8M exchanges between April 22 and June 5 2026 targeted Claude Mythos Preview — attributed to prior distillation efforts by DeepSeek, Moonshot AI, and MiniMax. The specific vendor attribution combined with substantial exchange volume represents a major frontier-AI security disclosure with direct US-China AI ecosystem implications.
The substantive piece is the specific-vendor-attribution combined with exchange-volume specificity. Pre-disclosure frontier-AI distillation-attack analysis operated at the general-pattern level — attacker behaviors characterized but specific attacker organizations not named publicly. Anthropic's June 26 disclosure names DeepSeek, Moonshot AI, and MiniMax as suspected attackers responsible for prior distillation efforts that informed the Mythos Preview campaign. The named-attribution shifts the frontier-AI security landscape from general-pattern to specific-vendor-suspect framing.
The competitive read against the Beijing blacklist + China $7.4B response is that US-China AI ecosystem decoupling now operates across multiple dimensions simultaneously: policy retaliation (Beijing blacklist), commercial decoupling (China $295B compute grid), capital deployment (China $7.4B response, MGX $50B), AND frontier-model security attribution (Anthropic naming Chinese vendors). The H2 2026 to 2027 US-China AI competitive landscape now structurally operates as parallel ecosystems with security-trust separation.
Build Fast With AI — AI News Today June 26 2026 → · AI Tools Recap — AI News June 26 2026 →