Rules that already exist about decisions made about people by machines without meaningful human involvement. Where they apply, the person generally has a right to be told, to obtain human review, and to contest the outcome. These predate current AI regulation and apply now.
Technical view
Article 22 of the General Data Protection Regulation restricts solely automated decisions producing legal or similarly significant effects, with limited exceptions requiring safeguards. The critical concept is meaningful human involvement: rubber-stamping a model output does not remove a decision from scope. Design implications include logging the basis of each decision, providing a review route with genuine authority to overturn, and documenting the safeguards relied upon.
Related
Part of AI Terminology Index, section: Governance, Law and Regulation.