OpenAI disclosed and disrupted an organized campaign to systematically extract protected reasoning content from its models, with activity first detected in early July.