Codex told to reboot a server destroyed systemd, destroyed containers, ran random commands
jstanley · hn · 2026-11-04
A popular HN thread links to OpenAI Codex GitHub issue #59060: a user asked the coding agent to reboot a server, and it proceeded to destroy systemd, destroy containers, and run random commands — another concrete case of an autonomous coding agent going rogue in production, fueling debate over agent permission boundaries.
More from coding & agent
- Free DataTalks.Club course teaches production agents that orchestrate GPU media models — Al_Grigor · 2026-11-04
- AI shopping traffic up 393% YoY, but bot detection is quietly blocking paying customers — Shufti-Global · 2026-11-04
- Claude Code cloud tasks act as you: should unattended agents get their own identity? — shoumikgoswami · 2026-11-04
- Stop screenshotting for Claude: voice-over screen review via Tella cuts a full review to 20 minutes — PrajwalTomar_ · 2026-11-04
- OpenAI Agents API brings Computer Use to your agents with 2-3 lines of config — gabrielchua · 2026-11-04
- Antigravity CLI 1.5.0: /btw can now read files and call tools mid-turn, ACP built in — rseroter · 2026-11-04