Compliance

special-purpose AI

AI systems whose outputs materially affect core human interests.

Definition

A category of AI systems identified in the draft that perform functions (learning, inference, perception, judgment, or language understanding) whose outputs can materially affect life, bodily safety, dignity, or other core interests; classification into this category triggers heightened duties such as prior-notification/registration, notice to affected persons, reporting, and explanation obligations. Precise technical thresholds and criteria for designation are delegated to subordinate implementing rules and ministerial guidance.