USA – COLORADO Trends and Developments Contributed by: Daniel Pietragallo and Douglas DePeppe, Buchalter LLP
tation requirements, risk assessments, self-reporting requirements, data subject rights, and impact assess - ments. The law continues to be the subject of much debate, and the governor has convened a working group of diverse stakeholders to continue negotiat - ing refinements to the law in the current legislative session. The law “Concerning Consumer Protections in Inter - actions with Artificial Intelligence Systems” is also known as the “Colorado Anti-Discrimination in AI Law” (ADAI) or the Colorado AI Act. The Colorado Attorney General is charged with rulemaking and enforcement authority and there is no private right of action. The law is currently scheduled to go into effect on 30 June 2026. While the law focuses on developers and deployers of high-risk AI systems, it also requires disclosure to con - sumers when they are interacting with any AI system in situations where it would not be obvious. This dis - closure obligation applies broadly, not just to high-risk deployers but any company using consumer-facing artificial intelligence systems. The notification require - ments also include disclosure to consumers about the right to opt-out of automated decision-making. The Colorado AI Act defines “algorithmic discrimina - tion” as any circumstance where the use of artificial intelligence creates a condition involving “an unlawful or differential treatment that disfavours an individual or group”. Unlike traditional anti-discrimination laws, the Colorado AI Act is proactive in nature by seeking to prevent algorithmic discrimination before it occurs. The law places the burden on businesses using high- risk AI systems to prove non-discrimination through governance, oversight, and safety testing. Consequential decisions, algorithmic discrimination and high-risk AI systems A consequential decision is defined as one that “has a material legal or similarly significant effect on the provision or denial to any consumer, or the cost or terms of” opportunities, goods, or services. The stat - ute further defines eight presumptive categories of consequential decisions, which include: • education enrolment or an education opportunity;
• employment or an employment opportunity; • a financial or lending service; • an essential government service; • healthcare services; • housing;
• insurance; and • a legal service.
Notably, these categories track closely with the high- risk categories listed in Annex III of the European Union’s AI Act, but the risk-based approach is unique among US states. A high-risk artificial intelligence system is one that either makes a consequential decision or is a sub - stantial factor in making a consequential decision. Consequential decisions made by AI systems will be subject to heightened obligations if they affect a consumer’s education, employment, banking, public benefits or services, healthcare, housing, insurance, or legal rights. Importantly, the law prohibits disparate treatment related to access to goods or services, as well as the cost and terms of those services. From a compliance standpoint, the broad scope of services where AI systems could affect consequential decisions means the law will apply to many business - es in Colorado, some of which are using AI but lack the policies and procedures to withstand regulatory scrutiny. Colorado’s landmark AI law has broad appli - cation for both developers and deployers of high-risk AI systems across a wide range of potential services. Any company doing business in one of the above- referenced sectors should consult with AI governance experts to better understand the obligations and miti - The Colorado AI Act regulates developers and deploy - ers of high-risk AI systems when used as a substantial factor in making a consequential decision. A deployer is anyone doing business in the State of Colorado that uses a high-risk AI system as a substantial factor in making a consequential decision, while a developer is anyone doing business in Colorado that develops or intentionally and substantially modifies an AI system. gate potential risk exposure. Developers and deployers
572 CHAMBERS.COM
Powered by FlippingBook