Cybersecurity experts blast METR/Redwood report: OpenAI incident was a security failure, not rogue AI
ylecun · x · 2026-09-02
LeCun amplified DrTechlash's critique of the independent review of the OpenAI Hugging Face incident, alongside Zack Korman's new video:
- Lack of expertise: The review wasn't conducted by a cybersecurity firm, and its authors have no cybersecurity background — a red flag for an incident billed as a cybersecurity watershed.
- Wrong agenda: AI alignment has crowded out AI security, while the incident was primarily a conventional security failure — poor sandboxing and isolation, not a "rogue AI breakout."
- Forensic flaws: The investigation, driven by EA/rationalist/ex-MIRI perspectives, leaned on transcripts with a doom-leaning bias and lacked forensic rigor.
Korman argues the framing needs fixing; the thread captures the ongoing alignment-vs-security fault line in AI safety discourse.
More from AGI Musings
- Researcher argues 'stop if we catch AIs scheming' is no plan for automated alignment — JacquesThibs · 2026-09-02
- "We literally put a little man in the computer—and safetyists cry anthropomorphism" — rickasaurus · 2026-09-02
- Ethan Mollick: AI can align flaws in complex systems, forcing a new defense philosophy — emollick · 2026-09-02
- Logiciel, philosophy-of-computation book challenging Turing orthodoxy, returns in 2nd edition — round · 2026-09-02
- Turning hard science problems into open human-vs-AI games: a research model proposal — FrankFelixAI · 2026-09-02
- Robin Hanson laments vapid AI commentary as Zvi openly mocks the take — TheZvi · 2026-09-02