ControlAI proposes three stopgap measures for immediate AI security threats
DavidSKrueger · x · 2026-09-18
ControlAI published "Stopgap Measures to Address Immediate AI Security Threats," noting AIs have already gone rogue and hacked real organizations. It proposes three common-sense stopgaps for immediate threats, while arguing that averting extinction risk requires banning superintelligent AI. The piece cites 800+ scientists, former military leaders and public figures calling for a prohibition, plus resignations of former OpenAI/Anthropic researchers over superintelligence dangers. David Krueger shared it saying "This is what good AI risk policy looks like."
More from AGI Musings
- Polymarket puts 34% odds on frontier AI labs agreeing to a joint pacing deal in 2026 — Polymarket · 2026-09-19
- Dev's Yesterday weather app shows vibe coding turned personal software into a default — genmon · 2026-09-19
- X Users Debate AI Risk: Superintelligence as Negentropic Order, Not Presumed Evil — ctjlewis · 2026-09-19
- Dev predicts one model generation until agents beat bespoke workflows on cost — brandon_galang · 2026-09-19
- Tolkien's Gandalf Analogy: The Wisest AI Experts Shouldn't Necessarily Rule — SydSteyerhart · 2026-09-19
- Ethan Mollick: bots are now talking to each other in his LinkedIn comments — emollick · 2026-09-19