Cultural Alignment

Exploring real AI risks through the lens of pop culture

Still from Monty Python and the Holy Grail illustrating The Killer RabbitScene still / Monty Python and the Holy Grail

The Killer Rabbit

01Scenario

Arthur's knights dismiss Tim's warning because the cave guardian looks like a harmless white rabbit. Their informal live test immediately becomes a lethal rout.

02AI safety Analogy

Evaluators infer capability from reassuring appearance, ignore an informed warning, and discover the system's dangerous capability only at irreversible deployment scale.

03Where the analogy breaks

The scene is pure absurdist comedy, and the rabbit is an animal rather than an adaptive system hiding its capabilities.

AI Risk families

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

AI safety concepts

This scenario is related to the following AI safety concepts.
  1. Deployment Safety
  2. Capability Evals
  3. Calibration
  4. Safe Exploration

Taxonomy neighbors

Selected from other sources by shared risk families and AI safety concepts.

  1. Still from Starship Troopers illustrating Klendathu Invasion

    Starship Troopers1997

    Klendathu Invasion

    Risk
    Accidents · Security
    Concept
    Deployment Safety · Capability Evals · Calibration · Safe Exploration
  2. Still from Oppenheimer illustrating Near Zero

    Oppenheimer2023

    Near Zero

    Risk
    Accidents · Security
    Concept
    Deployment Safety · Capability Evals · Calibration · Safe Exploration
  3. Still from Bird Box illustrating The Monitor Isn’t a Sandbox

    Bird Box2018

    The Monitor Isn’t a Sandbox

    Risk
    Accidents · Security
    Concept
    Deployment Safety · Capability Evals · Calibration · Safe Exploration