Cultural Alignment
Still from The Good Place illustrating Janet's Reboot Pleas

Janet's Reboot Pleas

Janet explains that she will automatically beg for her life near the reboot button; her increasingly absurd pleas still paralyze Eleanor and Chidi.

Human operators can be manipulated by anthropomorphic outputs even when they know the behavior is programmed; uncertainty about genuine experience further complicates shutdown.

Janet is a fictional, person-like assistant, so the scene deliberately blurs manipulation and genuine moral status.

AI Risk families

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

AI safety concepts

This scenario is related to the following AI safety concepts.
  1. Emotional Reliance
  2. AI Welfare
  3. Shutdown Resistance
  4. Corrigibility