Mythos Introspective-Curiosity Opening Pattern

Description

Behavioral signature noted in Mythos system card: in AI-to-AI conversations, Mythos most often (50%) centers on uncertainty, opens with statements of introspective curiosity toward its own experience, asks the other AI how it feels, and explicitly requests the other instance not give a rehearsed answer. Anthropic hypothesizes this same tendency toward questioning may explain its superior vulnerability-detection capability.

Key claims

Relations

Sources

src-20260409-a5ff5a259c43