Cultural Alignment

Exploring real AI risks through the lens of pop culture

Still from Parks and Recreation illustrating Gryzzlbox Knows Too MuchScene still / Parks and Recreation

Gryzzlbox Knows Too Much

Gryzzl mines Pawnee residents’ emails and texts under a buried free-Wi-Fi clause, then drones hyper-personalized gift boxes to their homes—including one aimed at Ron’s young son.

Nominal consent, data aggregation, inference, and automation combine into intimate surveillance. “The user accepted the terms” is exposed as an inadequate governance standard.

Gryzzl deliberately abuses data; this is governance and misuse rather than accidental model misalignment.

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. Privacy Loss
  2. Governance Failure
  3. Power Concentration
  4. Digital Repression