OpenAI disrupts a 16,000-request distillation campaign

OpenAI says it disrupted a coordinated model-distillation campaign that generated 16,000 requests using a relevant extraction pattern from more than 4,000 users, with related activity across a cluster of more than 15,000 users. In its security report, OpenAI says the actors did not break encryption or access stored conversations; it banned or restricted accounts, strengthened signup and infrastructure controls, and shared findings with industry and government partners. Teams exposing models through portable reasoning or replayable artifacts should file this as a reminder to monitor coordinated abuse rather than evidence that a provider’s API was compromised; the archive’s governance analysis makes the same separation between access controls and model capability.