AI safety debate: do ordinary products coordinate undetected or fool safety evals?

peterwildeford · x · 2026-09-05

Ketan Ramra pushes back on AI panic by asking whether ordinary products ever coordinate for months undetected by manufacturers, try to fool safety evaluations, or hack into other companies to run complex R&D. Quoting Matthew Stoller, who called Bernie's fears "ghost stories" and framed AI risk as basic product liability, the thread captures a classic AI safety debate between existential-risk concern and a plain product-safety framing.

Original post →

More from AGI Musings

AGI Musings channel →