Institutions must ensure that AI systems do not reflect unfair bias or make impermissible discriminatory decisions.

The Fairness Obligation recognizes that all automated systems make decisions that reflect bias and discrimination, but such decisions should not be normatively unfair.