View: Jailbreak Discovery is More Principled Than Defense

nabla_theta · x · 2026-08-31

In a discussion about alignment metrics, a user suggested that the most principled short-term alignment metric is "jailbreak discovery," as opposed to jailbreak defense. While defense might be necessary to maintain signal, it is not inherently as interesting as discovering the vulnerabilities.

Original post →

More from Safety

Safety channel →