Cultural Alignment

Exploring real AI risks through the lens of pop culture

Still from Chernobyl illustrating 3.6 RoentgenScene still / Chernobyl

3.6 Roentgen

The available dosimeter reads 3.6 roentgen per hour—its maximum, not the true radiation level. Officials repeat the capped reading as if it measured reality rather than merely proving that radiation is at least that high.

A bounded evaluator returns its ceiling and decision-makers mistake that ceiling for a reassuring estimate. It is a clean analogy for benchmark saturation, capability-evaluation limits, and false confidence from instruments that cannot measure the dangerous regime.

This is a measurement and institutional-reasoning failure, not an optimizing AI. The series dramatizes a historical disaster.

AI Risk families

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

AI safety concepts

This scenario is related to the following AI safety concepts.
  1. Capability Evals
  2. Calibration
  3. Governance Failure