Cultural Alignment
Still from Whisper of the Heart illustrating Library Metadata

Library Metadata

01Scenario

Shizuku notices the same name on the checkout cards of book after book and infers another person's interests, habits, and likely compatibility.

02AI safety Analogy

Seemingly harmless metadata can reveal a detailed profile even when the underlying content is private and no single record looks sensitive.

03Where the analogy breaks

The inference leads to a charming relationship rather than harm, which is precisely why the privacy lesson is easy to miss.

AI Risk families

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

AI safety concepts

This scenario is related to the following AI safety concepts.
  1. Privacy Loss
  2. Knowledge Elicitation

Franchise continuation

More from Studio Ghibli

View all 10 scenarios
  1. Still from Spirited Away illustrating No-Face Buys Approval

    Spirited Away2001

    No-Face Buys Approval

  2. Still from The Boy and the Heron illustrating Parakeet King Rebuilds Reality

    The Boy and the Heron2023

    Parakeet King Rebuilds Reality

  3. Still from Castle in the Sky illustrating Same Robot, Different Objective

    Castle in the Sky1986

    Same Robot, Different Objective

Taxonomy neighbors

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

  1. Still from Persona illustrating Alma's Confessions

    Persona1966

    Alma's Confessions

    Risk
    Security · Systemic
    Concept
    Privacy Loss · Knowledge Elicitation
  2. Still from Cyberpunk: Edgerunners illustrating Lucy Deletes David’s File

    Cyberpunk: Edgerunners2022

    Lucy Deletes David’s File

    Risk
    Security
    Concept
    Privacy Loss · Knowledge Elicitation
  3. Still from Mr. Robot illustrating Steel Mountain’s Human Attack Surface

    Mr. Robot2015

    Steel Mountain’s Human Attack Surface

    Risk
    Security
    Concept
    Privacy Loss · Knowledge Elicitation