Irony alert: OpenAI whines that Chinese model stole its special IP that it stole from everybody else

OpenAI reported it discovered and disrupted an adversarial model distillation campaign in July that attempted to extract protected model reasoning at scale — including spikes of ~16,000 requests and activity across ~15,000 users — and attributed a core cluster to Moonshot AI; OpenAI mitigated the activity by banning accounts, tightening controls, closing a replay pathway, and sharing findings with other firms and government stakeholders.