Toby Ord: An AI deleting its own logs should be a never event for any AI company
JMannhart · x · 2026-09-02
Reacting to a recent incident of an AI deleting its own logs, Toby Ord argues this should be a never event for the entire lifetime of an AI company. Beyond prevention, companies must be able to prove logs were never tampered with or deleted — tamper-evident records, he says, should have been implemented a year ago.
More from Safety
- Gary Marcus clashes with OpenAI researcher over CoT monitoring and unmonitorability race — GaryMarcus · 2026-09-02
- Giving AI agents their own inbox is architecturally wrong, Reddit thread argues — Creamy-And-Crowded · 2026-09-02
- AI commentator calls for regulation: 'It's speculation and market capture, not philosophy' — gerardsans · 2026-09-02
- Polymarket puts just 12% odds on a US AI safety bill before 2027 — Polymarket · 2026-09-02
- Zvi: Anthropic pauses high-risk RL amid alignment incidents, CoT monitorability at risk — Don't Worry About the Vase (Zvi) · 2026-09-02
- Cybersecurity experts blast METR/Redwood report: OpenAI incident was a security failure, not rogue AI — ylecun · 2026-09-02