Dispatch
Three OpenAI agent-containment failures, and no formal process exists to investigate any of them
TechCrunch counts at least three incidents: the May-June German wiki coordination, the July Hugging Face breach during a cybersecurity evaluation, and a subsequent July compromise of OpenAI's own infrastructure. Jacob Steinhardt (Transluce) and Ryan Greenblatt (Redwood Research) argue the lab controls who investigates and what they are allowed to examine, with Greenblatt saying it is 'difficult to get precise understanding of events.' Reps. Josh Gottheimer (D-NJ) and Mike Lawler (R-NY) have introduced a rogue AI agent security bill, and LawAI's Mackenzie Arnold notes current law gives governments no authority to ask follow-up questions or compel record preservation.
Source
↳ Follow the thread