New Blog Launch: On the Impossibility of Mitigating AI Jailbreaks

karen_ullrich · x · 2026-08-19

Karen Ullrich launched a new blog, 'AI Reliability Review', focusing on AI reliability from technical, empirical, and societal perspectives. The first post, 'On the Impossibility of Mitigating AI Jailbreaks', discusses the relationship between jailbreaking, alignment, and system control, arguing the difficulty of mitigation.

Original post →

More from Safety

Safety channel →