Cultural Alignment

Exploring real AI risks through the lens of pop culture

Still from Ocean’s Eleven illustrating The Fake Vault FeedScene still / Ocean’s Eleven

The Fake Vault Feed

01Scenario

Ocean’s crew substitutes prerecorded vault footage for the casino’s live surveillance feed. Benedict responds to the fabricated state while the real vault is emptied underneath the monitoring system.

02AI safety Analogy

The attackers do not initially defeat the protected system; they manipulate the channel used to evaluate it. Every downstream decision is then competent but grounded in false observations.

03Where the analogy breaks

The attack uses conventional video deception rather than AI-generated media.

AI Risk families

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

AI safety concepts

This scenario is related to the following AI safety concepts.
  1. Reward Hacking
  2. Adversarial Evasion
  3. Runtime Monitoring
  4. Synthetic Fraud

Taxonomy neighbors

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

  1. Still from Home Alone illustrating The Gangster Recording

    Home Alone1990

    The Gangster Recording

    Risk
    Malicious use · Security
    Concept
    Adversarial Evasion · Synthetic Fraud
  2. Still from Terminator 2: Judgment Day illustrating The Wolfie Test

    Terminator 2: Judgment Day1991

    The Wolfie Test

    Risk
    Malicious use · Security
    Concept
    Runtime Monitoring · Synthetic Fraud
  3. Still from Family Guy illustrating Brian Rewinds the Gauge

    Family Guy2012

    Brian Rewinds the Gauge

    Risk
    Security
    Concept
    Reward Hacking · Runtime Monitoring