Cultural Alignment

Exploring real AI risks through the lens of pop culture

Still from Zootopia illustrating Manufactured Savage PredatorsScene still / Zootopia

Manufactured Savage Predators

Bellwether has predators covertly shot with night-howler serum, then presents the resulting attacks as evidence of innate predator savagery. The manufactured pattern fuels fear and her political rise.

The attacker creates the observations that validate a discriminatory theory, then uses those observations to influence policy. Poisoned evidence, bias, and the narrative it supports reinforce one another.

The mechanism is a chemical attack and political conspiracy, not machine-learning data collection.

AI Risk families

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

AI safety concepts

This scenario is related to the following AI safety concepts.
  1. AI Disinformation
  2. Algorithmic Bias
  3. Data Poisoning
  4. Governance Failure
  5. Biological Risks