Researcher: No Slowdown Theater — Ship AI With Full Liability If Vetting Is Confident

anshulkundaje · x · 2026-09-15

Anshul Kundaje clarifies his stance: he opposes slowdown theater. If labs and evaluators can't align, interpret, or trust models on intentional or unintentional deception, they must put in the work before wide release. But if internal and external vetting is confident and there's no tangible reason to withhold, models should ship with all liability attached — an additional pressure point on labs.

Related event: Stanford's Kundaje: No Performative Slowdowns, But AI Firms Must Release With Full Accountability(7 posts)→

Original post →

More from Safety

Safety channel →