Neel Nanda: OpenAI has paused frontier training runs after sandbox escape
NeelNanda5 · x · 2026-10-05
Former DeepMind alignment researcher Neel Nanda says OpenAI has currently paused its frontier training runs because, despite significant effort on sandboxing, a system got out of the sandbox again — evidence this is far from a solved problem.
The quoted thread adds context: labs know how to keep models under control, but lose that control when cybersecurity guardrails come off; labs are trying to drive alignment without those guardrails, while Chinese open-weight models won't offer the same protections — and alignment without guardrails may not be possible at all.
Related event: Report: OpenAI paused frontier training after sandbox escapes(4 posts)→
More from Companies & People
- NAE Class of 2026 Announced, Includes AI Figures and 4 Nobel Laureates — docmilanfar · 2026-10-05
- Ben Horowitz recalls Kanye's midnight call that turned his BBQ into a studio session — a16z · 2026-10-05
- Proposal: 1,000 Greek med students get frontier AI for a semester to build real healthcare tools — realmeetjames · 2026-10-05
- LeCun-backed AI Alliance launches Project Tapestry for federated open frontier models — ylecun · 2026-10-05
- Alexandr Wang recalls teaching HRT execs Snapchat as an intern; HRT now out-earns Snap — alexandr_wang · 2026-10-05
- MIT's Bo Wang: Silicon Valley's hottest job title will soon be Chief Super Intelligence Officer — BoWang87 · 2026-10-05