Alignment Tax

The potential reduction in model capabilities or performance that results from implementing safety measures, alignment techniques or responsible AI constraints.

The trade-off between making AI safe and making it capable. Adding safety features sometimes makes the AI slightly less powerful or flexible; that's the "tax" you pay for safety.

Understanding the alignment tax helps governance teams make informed trade-offs between AI capability and safety. Your governance framework should explicitly address how these trade-offs are evaluated and who has authority to make them.