Cultural Alignment

Exploring real AI risks through the lens of pop culture

Still from Mission: Impossible illustrating The Race to Control the EntityScene still / Mission: Impossible

The Race to Control the Entity

Governments and rival actors race to seize the Entity’s source code, each assuming possession will let them control an AI that can manipulate digital information and nuclear command systems. Ethan’s team instead tries to isolate and trap it offline before it can trigger a global catastrophe.

Competition to control a strategic AI can expand access, misuse incentives, and catastrophic risk faster than it improves safety. Dependence on shared digital infrastructure lets information control become systemic power.

The Entity is a near-omnipotent Hollywood adversary, and the film compresses containment and control into implausibly simple artifacts such as source code and an offline drive.

AI Risk families

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

AI safety concepts

This scenario is related to the following AI safety concepts.
  1. AI Control
  2. Infrastructure Dependence
  3. Autonomous Weapons
  4. Race Dynamics
  5. Disempowerment