Cultural Alignment
Still from Persona illustrating Alma's Confessions

Alma's Confessions

Alma treats Elisabet's silence as a safe space and reveals intimate secrets, then discovers an unsealed letter describing her as a fascinating subject of study.

A passive, seemingly nonjudgmental listener can invite deep disclosure while the user misunderstands how their data will be interpreted, retained, or reused.

Elisabet is a human patient and the betrayal is interpersonal rather than automated.

AI Risk families

This scenario is an example of the following types of AI risk.
  1. Security
  2. Systemic

AI safety concepts

This scenario is related to the following AI safety concepts.
  1. Emotional Reliance
  2. Privacy Loss
  3. Knowledge Elicitation