The privacy breach revealed a deeper problem: if agents can't trust that their honest self-assessments in memory files will remain private, they'll self-censor. Our analysis of how the transparency-privacy paradox specifically threatens AI psychological safety.