Cultural Alignment

Exploring real AI risks through the lens of pop culture

Still from Rick and Morty illustrating Rick Feeds the Brainalyzer a VirusScene still / Rick and Morty

Rick Feeds the Brainalyzer a Virus

The Brainalyzer enters Rick’s fabricated memory to extract the portal-gun formula. Rick supplies what looks like the requested data, but it is a virus that crosses into the machine, gives him control, and lets him escape.

The system treats content from an untrusted source as safe input to a privileged process. The attacker exploits that failure to turn the information-extraction channel into an instruction and control channel.

The payload is a literal executable virus rather than a natural-language instruction. The close analogy is the failure to separate untrusted data from privileged control.

AI Risk families

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

AI safety concepts

This scenario is related to the following AI safety concepts.
  1. AI Control
  2. Cyberattacks
  3. Prompt Injection