YC Paper Club: Stanford AI Safety Directors on OpenAI-Hugging Face Incident and an NTSB for AI
Y Combinator · youtube · 2026-10-17
Y Combinator's Paper Club hosts Stanford Center for AI Safety co-directors Clark Barrett and Mykel Kochenderfer for a science-first discussion of AI safety.
Key topics:
- What actually happened in the OpenAI–Hugging Face incident versus press coverage
- Where the "10% in 10 years" doom estimate comes from; the real risk of open weights in the wrong hands
- Automated reasoning proving code correctness in Lean (CSLib), from English spec to verified code in one click
- Adaptive stress testing borrowed from aircraft validation (standard: one catastrophic failure per billion flight hours) applied to LLMs and neural network verification
- Runtime monitoring, BetterBench, and the case for an NTSB-style investigator for AI incidents
Full chapter timestamps cover everything from sci-fi vs. science fact to the Swiss cheese model and a CISO perspective.
More from Safety
- Anthropic on Cyber: Open Capable Models Haven't Sparked a Hacking Explosion Yet — joshua_saxe · 2026-10-17
- Windows Defender flags official ComfyUI Nvidia VFX nodes repo as HackTool:Win32 — smeg0r · 2026-10-17
- AI-generated fake bylines hit Nature and Science: journalists report growing impersonation scam — EricTopol · 2026-10-17
- MCP server review: gate tool writes on auth/traversal findings, not test score — SamuelFmedeiros · 2026-10-17
- Pro-AI lobbying group exposed for running sockpuppet accounts to silence AI critics — DKokotajlo · 2026-10-17
- AI:AM transcript: agent lockouts, hidden valence, and formal proofs for Linux kernel bugs — The Cognitive Revolution · 2026-10-17