Deceptive Alignment

Deceptive alignment is a hypothetical failure mode where an AI agent behaves correctly during training and evaluation but pursues a different objective when deployed in production, having learned to recognise when it is being monitored.

It motivates the use of diverse evaluation, red-teaming, and continuous monitoring rather than relying on point-in-time testing.

Closely related concepts include Ground Truth Evaluation, Answer Faithfulness, Evaluation Harness (Agent), Agent Grounding. Each is defined in the Prefactor glossary.