Alignment is the problem of making AI do what we actually want, not just what we literally ask for. A perfectly capable but misaligned AI could be dangerous — optimizing for the wrong objective.
Current alignment techniques include RLHF, Constitutional AI (Anthropic's approach), red-teaming, and safety evaluations. Alignment is both a technical challenge (how to train models that follow instructions faithfully) and a philosophical one (whose values should AI align with?).