AI Labs Don't Share Safety Research. Here's a Fix.

Every frontier AI lab is working on the same unsolved problem: how to keep AI systems that may one day be smarter than us honest, safe and under control. No law says they have to share what they learn. This video explains why that matters and walks through a proposal for a National Framework for Shared AI Alignment Research. Chapters 0:00 The breakout 0:32 Nobody has to share 0:55 The hardest exam ever written 1:24 What alignment research is 1:41 The exam is getting harder 2:15 Warning signs 2:40 Why labs keep it in the drawer 3:11 One test, thirteen models 3:40 The seatbelt story 4:00 It has been done before 4:29 The proposal 4:39 Part 1: A shared vault 5:04 Part 2: Someone to run it 5:27 Part 3: Referees 5:47 Part 4: Know who is building 6:11 What it does not do 6:44 Why it matters Email frontieralignment@gmail.com for links and/or sources.