OpenAI Discloses 6 AI Misalignment Cases — What They Reveal About Agent Controls

OpenAI disclosed six AI misalignment cases that highlight new risks around agent permissions, network access, and enterprise control boundaries.