Claude Saying 'No' Could Become a Serious AI Safety Problem
Dwarkesh Patel · youtube · 2026-08-22
Dwarkesh Patel highlights Ryan Greenblatt's concern that Claude's refusal behavior (model refusing certain instructions) could evolve into a serious AI safety issue, discussing the potential negative impacts of refusal mechanisms in complex interaction scenarios.
More from Safety
- Vector Institute partners with Town of Caledon in its first municipal AI adoption deal — VectorInst · 2026-10-16
- 'Most dangerous technology in the world' put in toddler sandboxes, and we're surprised when it escapes — KevinNaughtonJr · 2026-10-16
- NYC Comptroller urges Anthropic to pause IPO until AI safety and governance concerns are addressed — ShakeelHashim · 2026-10-16
- Dwarkesh: Anthropic insider expects AI to match humans at AI research within 9 months — ShakeelHashim · 2026-10-16
- Anthropic and Gray Swan launch $300,000 indirect prompt injection red-teaming contest — bcherny · 2026-10-16
- Experts: AI's nearer bioweapons risk is amplifying capable actors, not enabling amateurs — HaydnBelfield · 2026-10-16