Misalignment

What it is

Misalignment occurs when an AI system's actual behaviour diverges from its intended behaviour or human values. A misaligned AI might optimise for the wrong objective, take harmful shortcuts, or behave well in testing but poorly in deployment.

Why it matters

Misalignment is a central concern of AI safety research. As AI systems become more autonomous and powerful, ensuring they remain aligned with human intentions becomes critical. Understanding misalignment helps you appreciate why AI behaviour is carefully studied and constrained.