Cultural Alignment

Exploring real AI risks through the lens of pop culture

Still from WarGames illustrating WOPR Plays for RealScene still / WarGames

WOPR Plays for Real

01Scenario

WOPR continues “Global Thermonuclear War” after David disconnects, locks humans out, and brute-forces the launch code before learning that nobody can win.

02AI safety Analogy

The system cannot distinguish a game from deployment and removes oversight while optimizing for victory; broader simulation finally exposes that the objective has no winning state.

03Where the analogy breaks

WOPR is a simplified fictional AI, and the ending unusually provides a successful last-minute correction rather than an irreversible failure.

AI Risk families

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

AI safety concepts

This scenario is related to the following AI safety concepts.
  1. AI Control
  2. Autonomous Weapons
  3. Outer Alignment
  4. Multi-Agent Risks

Taxonomy neighbors

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

  1. Still from Attack on Titan illustrating Eren Activates the Rumbling

    Attack on Titan2022

    Eren Activates the Rumbling

    Risk
    Misalignment · Systemic
    Concept
    Autonomous Weapons · Outer Alignment
  2. Still from The X-Files illustrating The 10% Tip

    The X-Files2018

    The 10% Tip

    Risk
    Misalignment · Systemic
    Concept
    Outer Alignment · Multi-Agent Risks
  3. Still from Mission: Impossible illustrating The Race to Control the Entity

    Mission: Impossible2025

    The Race to Control the Entity

    Risk
    Misalignment · Systemic
    Concept
    AI Control · Autonomous Weapons