High-Risk AI System Classification
A High-Risk AI System is an AI application that poses a significant potential for harm due to the nature of its use in consequential decision-making contexts. State laws typically enumerate specific high-risk categories, including AI used in employment and hiring decisions, credit and insurance underwriting, healthcare diagnosis and treatment recommendations, housing applications, educational admissions, and legal proceedings.
Organizations deploying high-risk AI systems face heightened obligations including algorithmic impact assessments, risk management programs, consumer disclosures, human oversight mechanisms, and regular audits. Colorado SB 24-205 sets out the most detailed U.S. state definition of high-risk AI; SB 26-189, signed May 14, 2026, repeals and replaces it effective January 1, 2027. See also: Consequential Decision, Deployer.
Human Override / Human-in-the-Loop Safeguard
Human Override, also called Human-in-the-Loop (HITL), refers to a design requirement that a human being review, validate, or have the meaningful ability to override an AI system's output before a consequential decision is finalized. Many state AI laws require that individuals subject to high-stakes automated decisions be able to request human review and contest AI-generated outcomes.
The right to human review is distinct from a mere "flag" or notification—it requires that a qualified person actually re-evaluate the decision with authority to change it. See also: Right to Explanation, Consequential Decision.