OpenAI agent security engineer: alignment window is narrow, community must step up
Scobleizer · x · 2026-09-07
Scobleizer reshares the first tweet from joedaroo, an engineer working on Agent Security at OpenAI. He says what is coming is sobering, the window for alignment and monitorability is critically narrow, and everyone needs to level up. He describes the past few months of agent security work at OpenAI as extremely intense but notes colleagues across the lab take it seriously. He urges the AI community to spend less time arguing over who cares more about safety and more time working together, pointing readers to Jakub's post.
More from Companies & People
- Clara Shih on when students should start using AI: once you can judge the output — clarashih · 2026-09-07
- Claude Code's Boris Cherny: don't optimize token cost, maximize returns — rohanpaul_ai · 2026-09-07
- RLSlow team credited with inventing RL at scale for LLMs — morqon · 2026-09-07
- Paul Graham: Founders' strength comes from having experienced weakness — santoshpanda · 2026-09-07
- Ex-OpenAI researcher Lukasz Kaiser pens farewell to RLSlow reasoning team — RubenEVillegas · 2026-09-07
- AI researchers spend their days debugging 'incomprehensible' training code, engineer says — gabrielchua · 2026-09-07