Domingos: Public AI Jailbreak Exploits All Came From Vendors; Risk Lies in Access, Not Capability
Pedro Domingos noted that all publicly disclosed jailbreak exploits for powerful AI models have come from the vendors themselves. Commentators argue the real risk factor is not model capability but who can access it and what permissions it holds.
2026-10-04 ~ 2026-10-04 · 2 related posts
- Domingos: months of powerful AI models, yet all exploits came from the makers themselves — pmddomingos · 2026-10-04
- Commentary: AI risk hinges on permissions and system access, not raw model strength — AlexTensor · 2026-10-04