High-risk artificial intelligence system
An AI system that substantially factors in consequential decisions.
How the laws define this
Regulatory frameworks consistently anchor high-risk artificial intelligence systems to their role in making or substantially influencing consequential decisions that impact vital life outcomes, such as employment, healthcare, housing, and education. The critical divergence lies in how strictly instruments condition this status on human oversight, underlying technical architecture, and specific statutory carve-outs. The High-Risk Artificial Intelligence Developer and Deployer Act adopts a highly specific standard, restricting high-risk status to machine-learning-based systems operating without meaningful human oversight that autonomously dictate decisions about Virginians, while granting explicit exclusions for technologies like anti-fraud tools, networking, and narrow procedural tasks. By contrast, instruments including Colorado SB24-205, Connecticut SB 2, Increase Transparency for Algorithmic Systems, An act to amend the civil rights law and the executive law, and Concerning Consumer Protections in Interactions with Artificial Intelligence Systems apply broader definitions that encompass any system deployed as a substantial factor in consequential decisions without requiring a complete lack of human oversight or detailing sweeping technological exclusions.
Synthesised from the 6 statutory definitions below. Each one is quoted in full, with its source.
How jurisdictions differ
What is distinctive in each definition, binding law in force first.
| Jurisdiction | Instrument | Status | What is distinctive |
|---|---|---|---|
| Colorado, United States | Colorado SB24-205 — Consumer Protections for Artificial Intelligence Act | Repealed | Explicitly limits consequential decisions to material legal or similarly significant effects on a consumer's access to or conditions in specified sectors like education, employment, and healthcare. |
| New York, United States | An act to amend the civil rights law and the executive law, in relation to the use of artificial intelligence systems | Under Review | Broadens to include criminal justice and public services alongside standard sectors, and emphasizes decisions with significant effects on fundamental rights or opportunities. |
| Connecticut, United States | An Act Concerning Artificial Intelligence (Connecticut SB 2) | Withdrawn | Adds 'other similarly important life opportunities' beyond the listed sectors of education, employment, financial services, healthcare, housing, and insurance. |
| Virginia, United States | High-Risk Artificial Intelligence Developer and Deployer Act | Withdrawn | Requires machine-learning basis, no meaningful human oversight, principal basis for autonomous or substantially influential decisions, with numerous exclusions like anti-fraud tech and video games. |
| Colorado, United States | Increase Transparency for Algorithmic Systems | In Force | Omits 'substantial factor' phrasing, simply stating AI systems whose use in consequential decisions could have material legal or similarly significant effects. |
| Colorado, United States | Concerning Consumer Protections in Interactions with Artificial Intelligence Systems | Repealed | Limits to decisions related to employment, housing, credit, insurance, education, and healthcare where potential harm to consumers is greatest. |
What it triggers
- In Colorado, high-risk AI systems triggered enhanced compliance requirements including mitigation of algorithmic discrimination and impact assessments until the act was repealed.source
- Under Increase Transparency for Algorithmic Systems in Colorado, high-risk AI systems are subject to the delayed implementation of the artificial intelligence framework until mid-2026.source
- Under New York Senate Bill S1169, high-risk AI systems would require risk management programs for developers and deployers and prohibit algorithmic discrimination in key sectors.source
- In Colorado under SB24-205, high-risk AI systems required mitigation of algorithmic discrimination and impact assessments until it was repealed.
Definitions (6)
Any AI system that, when deployed, makes or is a substantial factor in making a consequential decision—i.e., a decision with material legal or similarly significant effects on a consumer's access to or conditions of education, employment, financial services, essential government services, healthcare, housing, insurance, or legal services. Such systems trigger the statute's enhanced compliance requirements.
An artificial intelligence system that, when deployed, makes, or is a substantial factor in making, a consequential decision. Consequential decisions are those pertaining to access to or the terms of education, employment, financial or lending services, healthcare services, housing, insurance, or other similarly important life opportunities.
Any AI system intended or reasonably foreseeable to be used to make consequential decisions that have a significant legal or similarly significant effect on an individual's fundamental rights or opportunities, including employment, education, housing, healthcare, financial services, insurance, criminal justice, and public services.
AI systems whose use in consequential decisions, such as those affecting employment, healthcare, housing, or government services, could have a material legal or similarly significant effect on individuals.
Refers to any AI system that, when deployed, makes or is a substantial factor in making a consequential decision. These decisions include those related to employment, housing, credit, insurance, education, and healthcare, where potential harm to consumers is greatest.
A machine-learning-based system that operates without meaningful human oversight, serves as the principal basis for, and is explicitly intended to autonomously make or substantially influence, consequential decisions about Virginians in individual or household contexts. Exclusions include systems for narrow procedural tasks, improving human activity, detecting patterns, or preparatory tasks, as well as anti-fraud technology (without facial recognition), AI-enabled video games, autonomous vehicle technology, cybersecurity technology, databases, data storage, firewall technology, networking, spam and robocall filtering, and natural language communicators (subject to an acceptable use policy).
Not to be confused with
High-risk automated decision system focuses on AI systems making decisions with significant legal impacts in sensitive domains, while high-risk artificial intelligence system definitions vary in specifying substantial factors, autonomy, or sectors.
High-Risk Artificial Intelligence broadly covers systems materially affecting life, safety, or fundamental rights, whereas high-risk artificial intelligence system often narrows to consequential decisions in sectors like employment or healthcare.
Questions and answers
What sectors count as consequential decisions for high-risk AI systems?
Most definitions focus on sectors like employment, education, healthcare, housing, financial services, and insurance. Some add criminal justice, public services, or other important life opportunities. Virginia narrows to individual or household contexts with exclusions.
Does high-risk AI include systems with human oversight?
Most definitions cover systems that make or are a substantial factor in decisions, without excluding human oversight. Virginia specifically requires operation without meaningful human oversight.
What obligations apply to high-risk AI in Colorado?
Colorado SB24-205 required mitigation of discrimination and impact assessments until repealed. SB25B-004 delays the AI framework implementation to mid-2026.
Related Terms
High-risk automated decision system
AI systems making decisions with significant legal impacts in sensitive domains....
High-risk AI applications
AI systems that significantly affect rights, safety, or essential services....
High‑Risk Artificial Intelligence
AI systems materially affecting life, safety, or fundamental rights....
High-Impact AI Systems
AI systems whose use poses significant risks to rights or safety....
High‑impact AI system
AI systems posing significant risk to people or society....