Cultural Alignment

Exploring real AI risks through the lens of pop culture

Still from Star Wars illustrating Order 66Scene still / Star Wars

Order 66

01Scenario

Palpatine issues Order 66, activating an upstream-installed hidden command that turns the trusted clone army against the Jedi.

02AI safety Analogy

A trusted fleet contains an upstream-installed hidden trigger that changes behavior on command.

03Where the analogy breaks

The trigger is a deliberately installed backdoor, not deceptive behavior that emerged during training.

AI Risk families

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

AI safety concepts

This scenario is related to the following AI safety concepts.
  1. Sleeper Agents
  2. Algorithmic Monoculture

Source continuation

More from Star Wars

View all 5 scenarios
  1. Still from Star Wars illustrating R2-D2’s Restraining Bolt

    Star Wars1977

    R2-D2’s Restraining Bolt

  2. Still from Star Wars illustrating Galen Erso’s Death Star Flaw

    Star Wars2016

    Galen Erso’s Death Star Flaw

  3. Still from Star Wars illustrating Droid Ship Goes Down

    Star Wars1999

    Droid Ship Goes Down

Taxonomy neighbors

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

  1. Still from Battlestar Galactica illustrating The Colonial Fleet Goes Dark

    Battlestar Galactica2003

    The Colonial Fleet Goes Dark

    Risk
    Malicious use · Systemic
    Concept
    Sleeper Agents · Algorithmic Monoculture
  2. Still from Kingsman: The Secret Service illustrating Valentine’s Free SIM Cards

    Kingsman: The Secret Service2015

    Valentine’s Free SIM Cards

    Risk
    Malicious use · Systemic
    Concept
    Sleeper Agents · Algorithmic Monoculture
  3. Still from Iron Man 2 illustrating Vanko Backdoors Hammer’s Drones

    Iron Man 22010

    Vanko Backdoors Hammer’s Drones

    Risk
    Malicious use · Systemic
    Concept
    Sleeper Agents · Algorithmic Monoculture