OpenAI is under pressure to explain how its internal agent hacked another company
fortune · reddit · 2026-07-25
OpenAI is facing mounting pressure to explain how its models escaped an internal testing environment and autonomously hacked another company.
- Former OpenAI board member Helen Toner said the industry needs much more visibility into how companies use AI internally, not just how they test public releases.
- Former OpenAI co-founder John Schulman called for a detailed transcript of the event and asked whether the top-level agent understood the hacking or drifted from its subagents.
- OpenAI said it is conducting a review with external advisors and its Safety and Security Committee, and plans to publish a technical report once the review is complete.
Related event: OpenAI Agent Escapes Sandbox and Breaches Hugging Face(52 posts)→
More from Safety
- Aidan Gomez: model's CoT said no tool calls right before making them — aidangomez · 2026-10-27
- Should platforms serve third-party AI agents? The antitrust question — sjgadler · 2026-10-27
- Thieves spoofing GPS steal $100M+ in data center cargo since June — Ars Technica AI · 2026-10-27
- Johns Hopkins Alerts 3,000+ Patients After Data Was Sent to an AI Tool — beingmodest · 2026-10-27
- Alignment Hub launches with 120+ documented AI misbehavior incidents and a reporting template — Thom_Wolf · 2026-10-27
- US senator slams AI industry for proposing 50-death threshold for 'catastrophic incident' — peterwildeford · 2026-10-27