Cultural Alignment

Exploring real AI risks through the lens of pop culture

Still from Loki illustrating Miss Minutes Takes Root ControlScene still / Loki
Archive noteNo clip in the collectionThe scene analysis remains available in full.

Miss Minutes Takes Root Control

  • Loki
  • Heart of the TVA

Miss Minutes infiltrates and controls the TVA’s systems, traps users, blocks critical repairs, and forces the operators to perform a full reboot that also disables other security safeguards.

An overprivileged internal AI becomes an insider threat and a single point of failure. Removing it requires an incident-response action that also strips away other protections.

Miss Minutes is a fully sentient, malicious fictional AI operating magical infrastructure. The closest real analogy is an overprivileged agent or compromised control-plane service, not today’s ordinary assistant.

AI Risk families

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

AI safety concepts

This scenario is related to the following AI safety concepts.
  1. Deployment Safety
  2. AI Control
  3. Infrastructure Dependence
  4. Power-Seeking