Cultural Alignment
Still from Akira illustrating SOL Only Takes an Arm

SOL Only Takes an Arm

The Colonel fires the orbital laser SOL as a last-resort attempt to stop Tetsuo. It severs one arm but fails to disable him; Tetsuo identifies the weapon, destroys the satellite, and later rebuilds his arm from machinery.

A control measure that is neither decisive nor robust can expose itself, provoke an adaptive target, and remove its own future usefulness. Last-resort shutdown mechanisms must be tested against the capability levels they are meant to contain.

SOL is a lethal attack on a human who is already openly hostile, not a normal model shutdown; provoking resistance is part of the scene’s causal story.

AI Risk families

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

AI safety concepts

This scenario is related to the following AI safety concepts.
  1. Deployment Safety
  2. Capability Evals
  3. AI Control
  4. Adversarial Robustness
  5. Shutdown Resistance