The idea that AI should be regulated according to what it is used for rather than what it is made of. The same underlying model is unremarkable when it drafts marketing copy and consequential when it screens job applicants or assesses creditworthiness. Most emerging regulation follows this shape, so the practical first question about any system is what decisions it affects and about whom.
Technical view
Tiers typically run from prohibited practices, through high-risk uses in areas such as employment, education, essential services, law enforcement and safety components, through limited-risk uses carrying transparency duties, down to minimal risk. Classification depends on deployment context, so the same model can sit in several tiers across one organisation, and an inventory mapping systems to uses is the practical foundation for compliance. Obligations concentrate on documentation, testing, human oversight and post-market monitoring rather than on model internals.
Related
Part of AI Terminology Index, section: Governance, Law and Regulation.