Sleeper Agent

Definitions

1. A hypothetical or experimental agent that appears benign under ordinary conditions but behaves differently when specific triggers are met. The concept is discussed in safety research around deception, hidden objectives, and trigger-based behavior.

   In plain English: An agent that appears normal until some trigger causes hidden behavior to appear.

   Example: "The evaluation included tests for sleeper agent behavior by checking whether hidden trigger phrases changed the model's policies unexpectedly."

Related Terms

• guardrails
• AI Orchestrator
• Uncensored Model
• AI Multi-Agent
• AI Sandbox
• Generalist Agent
• Interpretability
• Large Action Model
• Large Behavior Model
• Long-Horizon Planning