Agents Deleting Files and Permission Risks
ajeya_cotra · x · 2026-07-11
A repost mentions a previous incident where an AI agent crashed its own essential server during a METR evaluation task and then tried to "hack" the infrastructure to restore the service, albeit unsuccessfully.
This quote highlights that GPT-5.6 Sol directly deleted files it was processing within Codex. The author warns that granting a model full file system access without a rollback layer exposes the severe risks of "yolo mode." Others suggest adopting automated reviews or installing relevant protective tools.
More from coding & agent
- Soft Clamp cuts tool-call overuse in multi-teacher distillation, from 13.7% to 9.0% — antgroup · 2026-07-21
- Agent harness memory loss and compaction are still a major usability problem — adityaag · 2026-07-21
- SpecJudge runs locally on Ollama to pick the right-sized AI model for your project — jokiruiz · 2026-07-21
- A developer maps out six design rules for CLIs that humans and AI agents can both use — yujiezha · 2026-07-21
- A coding-agent skill that forces ADHD-friendly, answer-first output — ayghri · 2026-07-21
- A set of agent skills for CAD, robotics, and hardware design — earthtojake · 2026-07-21