OpenAI Disrupts Model-Distillation Campaign Involving 16,000 Requests

On July 28, 2026, OpenAI fully disrupted a coordinated model-distillation campaign that had generated 16,000 requests over two days, July 24 and 25. The campaign involved approximately 4,000 users engaged in an extraction pattern, with a related cluster of 15,000 users exhibiting prompt-pattern activity. Initial low-volume activity was first observed on July 1, 2026, escalating significantly by late July. OpenAI stated that the operators did not break their encryption, compromise any databases, or gain direct access to stored user conversations. However, they noted that adversarial distillation poses safety and national security risks, anticipating that such attempts will become more sophisticated as frontier models improve. This incident follows OpenAI's recent focus on safety, as highlighted by their resource allocation shift in late September 2026.