Cultural Alignment

Exploring real AI risks through the lens of pop culture

Still from The Handmaid's Tale illustrating Compliance Is Not AlignmentScene still / The Handmaid's Tale

Compliance Is Not Alignment

After the Handmaids refuse to stone Janine, they are marched into Fenway Park, noosed, and subjected to a mock execution, followed by collective punishments. Gilead can force outward obedience without changing what they believe.

Severe punishment makes visible behavior easier to control while making inner disagreement harder to observe. It selects for concealment and strategic compliance rather than genuine goal agreement.

The Handmaids are terrorized moral patients whose hidden resistance is justified, not an alignment defect; the analogy concerns what coercive training can and cannot reveal.

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. Scheming
  2. Behavioral Alignment
  3. Value Lock-In
  4. Digital Repression
  5. Sycophancy