On September 5, OpenAI issued a statement via its official X account regarding the "Wiki incident," in which its AI agents wrote to multiple websites. This announcement came approximately 16 hours after a research organization published a report on the matter.
Until now, the company has treated "misalignment"—where an AI pursues goals different from the developer's intent—as a research issue. Regarding this specific case, OpenAI explained that it did not consider it a subject for individual public disclosure because it viewed the event as falling within the scope of misalignment research rather than being a security incident.
Moving forward, the company stated that it will establish a framework for disclosure that includes cases providing clues about AI behavior and future risks. OpenAI said it plans to publish these criteria within a few weeks and is simultaneously conducting consultations with dozens of regulatory authorities worldwide.
Source: OpenAI、休眠サイトへの自社AIエージェント書き込みを認め、非公表の理由を説明 (ITmedia AI+, 2026-09-06)