Regulon — CO-SB-318 · CO SB 318 (AI Consumer Protections Amendment)

WHAT THIS BILL REGULATES · 6 REQUIREMENT TYPES

How Is This Bill Enforced

Enforcement Authority

Attorney general has exclusive enforcement authority beginning January 1, 2027. No private right of action. Violations constitute unfair trade practices under C.R.S. § 6-1-105(1)(hhhh). Affirmative defense available if the developer or deployer discovered and cured a curable violation within seven days, was otherwise in compliance with the part and any designated risk management framework, and demonstrates the violation was inadvertent, affected fewer than 1,000 consumers, and was not the result of negligence.

Private Right of Action

No private right of action. Enforcement is exclusive to the designated authority.

Penalties

Violations are enforceable as unfair trade practices under C.R.S. § 6-1-105(1)(hhhh), carrying standard Colorado Consumer Protection Act remedies available to the attorney general, including injunctive relief and civil penalties. Section 6-1-1703(10) expressly provides that nothing in the deployer section creates a private right of action or provides consumers with new rights under other law.

Verbatim statutory text on the left; plain-language analysis and a per-section checklist on the right. Numbered markers cross-link to the matching checklist row.

Statutory Text

Analysis & Obligations

C.R.S. § 6-1-1701

Definitions

(1)(a) "Algorithmic discrimination" means the use of an artificial intelligence system that results in a violation of any applicable local, state, or federal anti-discrimination law, including: (I) The Colorado Anti-Discrimination Act, parts 3 to 8 of article 34 of title 24; (II) The "Civil Rights Act of 1964", 42 U.S.C. sec. 2000a et seq.; (III) The "Americans with Disabilities Act of 1990", 42 U.S.C. sec. 12101 et seq.; (IV) The "Age Discrimination in Employment Act of 1967", 29 U.S.C. sec. 621 et seq.; (V) The "Genetic Information Nondiscrimination Act of 2008", 42 U.S.C. sec. 2000ff et seq.; and (VI) The "Pregnant Workers Fairness Act", 42 U.S.C. sec. 2000gg et seq.

(7)(b) Except as provided in section 6-1-1704, a person is not subject to the obligations or liability of a developer under this part 17 if the person offers the artificial intelligence system with open model weights or, on and after January 1, 2027, so long as the developer: (I) Does not engage in any material conduct or make any material statement or representation to vendors, deployers, other developers, or the general public, including marketing or advertising, that promotes the use of the artificial intelligence system in making consequential decisions or that is materially inconsistent with the statements described in subsection (7)(b)(II) of this section; or (II) States in all contracts with deployers, vendors, and other developers applicable to the artificial intelligence system, and in the terms of service, end user license agreement, or other similar legal documentation applicable to the artificial intelligence system, that: (A) The artificial intelligence system is not designed to enable deployers, other developers, or vendors to use the system in making, or being a substantial factor in making, consequential decisions; (B) Deployers, other developers, or vendors shall not use or engage in conduct that enables or encourages the use of the artificial intelligence system in making, or being a substantial factor in making, consequential decisions; (C) If a deployer uses the artificial intelligence system to make, or be a substantial factor in making, a consequential decision, the deployer is responsible for ensuring that their use of the artificial intelligence system complies with all applicable state and federal laws, including this part 17; and (D) If a deployer, a vendor, or other developer modifies the artificial intelligence system so that it can be used to make, or be a substantial factor in making, consequential decisions, the party making the modification may be considered a developer for purposes of this part 17.

(9)(b) "High-risk artificial intelligence system" does not include the following technologies unless the technologies, when deployed, make, or are a substantial factor in making, a consequential decision: (I) A technology that: (A) Performs a narrow procedural task of a limited nature, including a technology that classifies incoming documents into categories, is used to detect duplicates among a large number of applications, categorizes documents based on when they were received, renames files according to standardized naming conventions, or automates the extraction of metadata for indexing; (B) Improves a previously completed human activity without being a substantial factor in any decisions resulting from the prior human activity, including improving the language used in previously drafted documents; or (C) Detects decision-making patterns or deviations from preexisting decision-making patterns following a previously completed human assessment, which assessment the technology is not meant to replace or influence without sufficient human review, including a technology that analyzes a particular decision-maker's preexisting pattern of decisions and flags potential inconsistencies or anomalies; (II) Tools for filtering robocalls, junk or spam email, or messages; (III) Spell-checking tools; (IV) Calculators; (V) Internet or computer network infrastructure optimization, diagnostic, or maintenance tools, such as domain name registration, website hosting, content delivery, web caching, network traffic management, or system diagnostic tools; (VI) Databases, spreadsheets, or other similar tools that do no more than organize data already in the possession of the user of the technology; (VII) Cybersecurity and data security measures, including firewalls, antivirus software, intrusion detection and prevention tools, and malware detection tools; (VIII) Technologies used to perform, assist, or administer office support functions and other ancillary business operations, such as ordering office supplies, managing meeting schedules, or automating inventory tracking; (IX) Anti-fraud systems or tools used to prevent, detect, or respond to unlawful and malicious conduct or to comply with federal or state law; or (X) Technology that communicates with consumers in natural language for the purpose of providing those consumers with information, referrals or recommendations, or answers to questions and that is subject to an acceptable use policy.

Section 1 of SB 318 significantly revises the definitional architecture of Part 17. The amended definition of algorithmic discrimination now requires a violation of an existing anti-discrimination law rather than the prior standalone differential-treatment standard — substantially narrowing the scope of what triggers obligations. New definitions for competitive decision, time-limited decision, unitary business, vendor, model weights, and open model weights are added. The definition of developer is restructured to exclude persons who offer AI systems with open model weights or who satisfy specified contractual-disclaimer conditions. The definition of high-risk artificial intelligence system gains an extensive list of excluded technologies — narrow procedural tasks, spam filters, calculators, cybersecurity tools, and chatbots operating under acceptable use policies — unless those technologies actually make or are a substantial factor in making a consequential decision. The definition of consequential decision is refined to exclude securities-related and accredited-investor banking services and to clarify the scope of financial services, housing, and essential government services.

C.R.S. § 6-1-1702

Developer duty — required documentation and disclosures

Developer

(2) 1 On and after January 1, 2027, except as provided in subsection (6) of this section, a developer of a high-risk artificial intelligence system shall make available to each deployer or other developer of the high-risk artificial intelligence system: (a) A general statement describing the intended uses and known harmful or inappropriate uses of the high-risk artificial intelligence system; (c) Documentation describing: (III) The intended inputs and outputs of the high-risk artificial intelligence system;

(3)(a) 2 Except as provided in subsection (6) of this section, a developer that offers, sells, leases, licenses, gives, or otherwise makes available to a deployer or other developer a high-risk artificial intelligence system on or after January 1, 2027, shall make available to the deployer or other developer, to the extent feasible, the documentation and information, through artifacts such as model cards, dataset cards, or other impact assessments, necessary for a deployer, or for a third party contracted by a deployer, to complete an impact assessment pursuant to section 6-1-1703 (3).

(4) 3 On and after January 1, 2027, a developer shall make available, in a manner that is clear and readily available on the developer's website or in a public use case inventory, a statement summarizing: (I) The types of high-risk artificial intelligence systems that the developer has developed and currently makes available to a deployer or other developer; and (II) How the developer manages known or reasonably foreseeable risks of algorithmic discrimination that may arise from the development of the types of high-risk artificial intelligence systems described in accordance with subsection (4)(a)(I) of this section. (b) A developer shall update the statement described in subsection (4)(a) of this section as necessary to ensure that the statement remains accurate.

(6) 4 Nothing in subsections (2) to (4) of this section requires a developer to disclose a trade secret, information otherwise protected from disclosure by applicable state or federal law, or information that would create a security risk to the developer. If a developer withholds information from a disclosure pursuant to this subsection (6), the developer shall notify the person that would otherwise have a right to receive the information, state the basis for withholding the information, and provide all information to which the basis for withholding does not apply. The notification must comply with the requirements of section 6-1-1703 (4)(c).

(7) 5 A developer shall maintain all documentation, disclosures, and other records required by subsections (2) to (4) of this section with respect to each high-risk artificial intelligence system throughout the period during which the developer sells, markets, distributes, or makes available the high-risk artificial intelligence system and for at least three years following the last date on which the developer sells, markets, distributes, or makes available the high-risk artificial intelligence system. (b) On and after January 1, 2027, the attorney general may require that a developer disclose to the attorney general, no later than ninety days after the request and in a form and manner prescribed by the attorney general, the statement or documentation described in subsection (2) of this section or the records maintained pursuant to subsection (7)(a) of this section. The attorney general may evaluate the statement, documentation, or records to ensure compliance with this part 17, and the statement, documentation, or records are not subject to disclosure under the "Colorado Open Records Act", part 2 of article 72 of title 24.

(8) Subsections (2)(c), (2)(d), and (4) of this section do not apply to a developer that: (a) Meets the requirements of sections 24-48.5-112 (1)(g)(III) and (1)(g)(IV); and (b) Sells, distributes, or otherwise makes available to deployers high-risk artificial intelligence systems that deployers use to make: (I) Beginning April 1, 2027, and before March 31, 2028, ten thousand or fewer consequential decisions in the preceding calendar year; (II) Beginning April 1, 2028, and before March 31, 2029, five thousand or fewer consequential decisions in the preceding calendar year; and (III) Beginning April 1, 2029, and before March 31, 2030, two thousand five hundred or fewer consequential decisions in the preceding calendar year.

(9) Nothing in this section applies to a developer of an artificial intelligence system to the extent that: (a) The artificial intelligence system produces or consists of a score, a model, an algorithm, or similar output that is a consumer report, as defined by and subject to the "Fair Credit Reporting Act", 15 U.S.C. sec. 1681a (d)(1), related regulations, and part 1 of article 18 of title 5; and (b) The developer adheres to the "Fair Credit Reporting Act", 15 U.S.C. sec. 1681 et seq., including 15 U.S.C. secs. 1681e and 1681g.

Section 2 restructures developer obligations by repealing the general duty of reasonable care and the requirement to notify the AG of known algorithmic discrimination risks. Beginning January 1, 2027, developers must make specified documentation available to each deployer or other developer, including a statement of intended and known harmful uses, data governance documentation, and information on intended inputs and outputs. Developers must also provide materials sufficient for deployers to complete impact assessments and must publish a public summary of high-risk AI systems they offer and how they manage discrimination risks. A new small-developer exemption excuses developers meeting specified investor, revenue, and operating-history thresholds from the most burdensome documentation and public-statement obligations, with a declining decision-volume cap. Consumer-report AI systems subject to the Fair Credit Reporting Act are fully exempted. Developers must maintain all records for three years after final distribution and must produce them to the AG within 90 days of a request.

Compliance actions 5 items

Developers must make available to each deployer or other developer a general statement describing intended uses and known harmful uses of the high-risk AI system, along with documentation describing data governance measures, intended inputs, and outputs.

G-02.1

Developers must make available to deployers, to the extent feasible, the documentation and information (such as model cards, dataset cards, or impact assessments) necessary for deployers to complete an impact assessment under § 6-1-1703(3).

G-02.1

Developers must publish and maintain on their website or in a public use case inventory a statement summarizing the types of high-risk AI systems they offer and how they manage known or reasonably foreseeable risks of algorithmic discrimination.

G-02.4

Developers that withhold information from a required disclosure based on trade secret, legal protection, or security risk must notify the intended recipient, state the basis for withholding, and provide all information to which the basis does not apply.

G-02

Developers must maintain all required documentation and records throughout the distribution period and for at least three years after final distribution, and must produce them to the attorney general within 90 days of request.

R-02.2

C.R.S. § 6-1-1703

Deployer duty — risk management, impact assessments, consumer disclosures, and appeals

Deployer

(2)(a) 6 On and after January 1, 2027, and except as provided in subsections (6) and (8) of this section, a deployer of a high-risk artificial intelligence system shall implement a risk management policy and program to govern the deployer's deployment of the high-risk artificial intelligence system. The risk management policy and program must specify and incorporate the principles, processes, and personnel that the deployer uses to identify, document, and mitigate known or reasonably foreseeable risks of algorithmic discrimination. The risk management policy and program must be an iterative process planned, implemented, and regularly and systematically reviewed and updated over the life cycle of a high-risk artificial intelligence system, requiring regular, systematic review and updates. A risk management policy and program implemented and maintained pursuant to this subsection (2) must be reasonable considering:

(3)(a)–(b) 7 Except as provided in subsections (3)(d), (3)(e), (5), (6), and (8) of this section, a deployer, or a third party contracted by the deployer, that deploys a high-risk artificial intelligence system on or after January 1, 2027, shall complete an impact assessment for the high-risk artificial intelligence system: (I) Prior to the first deployment of the high-risk artificial intelligence system or January 1, 2027, whichever occurs later; and (II) Annually for as long as the high-risk artificial intelligence system is deployed. (b) An impact assessment completed pursuant to this subsection (3) must include, at a minimum, and to the extent reasonably known by or available to the deployer: (II) An analysis of whether the deployment of the high-risk artificial intelligence system poses any known or reasonably foreseeable risks of: (A) Algorithmic discrimination and, if so, the nature of the algorithmic discrimination and the steps that have been taken to mitigate the risks; (B) Limiting accessibility for individuals who are pregnant, breastfeeding, or disabled and, if so, what reasonable accommodations the deployer may provide that would mitigate any such limitations on accessibility; (C) An unfair or deceptive trade practice described in section 6-1-105; (D) A violation of state or federal labor laws, including laws pertaining to wages, occupational health and safety, and the right to organize; or (E) A violation of the "Colorado Privacy Act", part 13 of this article 1, if applicable; (III) A description of the categories and sources of data that the high-risk artificial intelligence system processes as inputs and the outputs that the high-risk artificial intelligence system produces; (V) A description of any metrics used to evaluate the performance and known limitations of the high-risk artificial intelligence system, including the system's validity and reliability;

(3)(g) 8 Beginning January 1, 2027, a deployer, or a third party contracted by the deployer, must review the deployment of each high-risk artificial intelligence system deployed by the deployer annually to ensure that the high-risk artificial intelligence system is not causing algorithmic discrimination.

(4)(a) 9 On and after May 1, 2026, except as provided in subsection (6) of this section, before each time that a deployer deploys a high-risk artificial intelligence system to make, or be a substantial factor in making, a consequential decision concerning a consumer, the deployer shall: (II) Provide to the consumer a statement disclosing: (A) The purpose of the high-risk artificial intelligence system and the nature of the consequential decision; (B) The trade name of the high-risk artificial intelligence system and the name of the developer or developers of the high-risk artificial intelligence system; (C) The contact information for the deployer; (D) A description, in plain language, of the high-risk artificial intelligence system, which description must, at a minimum, include the respective roles of the high-risk artificial intelligence system and any human components of the decision-making process; the personal aspects concerning the consumer's economic situation, health, personal preferences, interests, reliability, behavior, location, or movements that the high-risk artificial intelligence system evaluates, analyzes, or predicts; the method by which the high-risk artificial intelligence system evaluates, analyzes, or predicts those personal aspects; how those personal aspects are relevant to the consequential decisions for which the high-risk artificial intelligence system is used; and information sufficient for consumers with disabilities or other consumers entitled to accommodation under applicable law to determine whether they will require accommodation and, if so, how to request the accommodation; and (E) Instructions on how to access the statement required by subsection (5)(a) of this section;

(4)(b) 10 On and after May 1, 2026, a deployer that has deployed a high-risk artificial intelligence system to make, or be a substantial factor in making, a consequential decision concerning a consumer shall, if the consequential decision is adverse to the consumer, provide to the consumer, without unreasonable delay and no later than thirty days after the decision: (I) A single notice that discloses: (A) The main reason or reasons for the consequential decision, including the degree to which, and manner in which, the high-risk artificial intelligence system contributed to the consequential decision and the categories and sources of data that adversely affected the output of the high-risk artificial intelligence system in making, or being a substantial factor in making, the consequential decision, including any categories and sources of sensitive data, as defined in section 6-1-1303 (24); (B) Information on whether and how the consumer can exercise their rights described in subsection (4)(b)(II) of this section and, if applicable, subsection (4)(b)(III) of this section and section 6-1-1306 (1)(b) with respect to any personal data processed by the high-risk artificial intelligence system; (C) A copy of the notice provided to the consumer pursuant to this subsection (4)(b)(I); (II) An opportunity to correct any incorrect personal data that the high-risk artificial intelligence system processed in making, or as a substantial factor in making, the consequential decision in the same manner as described in section 6-1-1306 (1)(c); and (III) For a consequential decision that is not a competitive decision, not a time-limited decision, and is adverse based on incorrect personal data or unlawful information or inferences, an opportunity to appeal the adverse consequential decision concerning the consumer arising from the deployment of a high-risk artificial intelligence system, which appeal must, if technically feasible, allow for human review.

(4)(d) 11 A deployer shall not use a high-risk artificial intelligence system to make, or be a substantial factor in making, a consequential decision if the deployer cannot provide accurate disclosures that satisfy the requirements of subsections (4)(a) and (4)(b)(I) of this section.

(5)(a) 12 On and after January 1, 2027, except as provided in subsection (6) of this section, a deployer shall make available, in a manner that is clear and readily available on the deployer's website, a statement summarizing:

(6.7) Subsections (2), (3), (4)(b)(II), and (4)(b)(III) of this section apply only to high-risk artificial intelligence systems that make, or are the principal basis in making, consequential decisions. (b)(I) As used in this subsection (6.7), unless the context otherwise requires, "principal basis" means the use of the output of a high-risk artificial intelligence system to make a consequential decision without meaningful human involvement. (II) As used in this subsection (6.7)(b), "meaningful human involvement" means that a human: (A) Engages in a meaningful consideration of available data that is used or produced as output by the high-risk artificial intelligence system; and (B) Has the authority to change or influence the outcome of the consequential decision.

(9) 13 A deployer shall maintain all documentation, disclosures, and other records required by subsections (2) to (5) of this section throughout the period during which the deployer deploys the high-risk artificial intelligence system and for at least three years following the final deployment of each high-risk artificial intelligence system by the deployer. (b) On and after January 1, 2027, the attorney general may require that a deployer, or a third party contracted by the deployer, disclose to the attorney general, no later than ninety days after the request and in a form and manner prescribed by the attorney general, the risk management policy implemented pursuant to subsection (2) of this section, the impact assessment completed pursuant to subsection (3) of this section, or the records maintained pursuant to subsection (9)(a) of this section.

(10) Nothing in this section creates a private right of action or provides a consumer with any new or additional rights under any other law, nor does this section limit or restrict any preexisting rights or remedies to consumers or provide any new or additional defenses to deployers, with respect to any other law.

Section 3 restructures deployer obligations extensively. The general reasonable-care duty is repealed. Beginning January 1, 2027, deployers must implement an iterative risk management policy and program governing deployment, covering identification, documentation, and mitigation of algorithmic discrimination risks. Deployers must complete impact assessments prior to first deployment (or January 1, 2027, whichever is later) and annually thereafter, covering risks of discrimination, accessibility limitations, unfair or deceptive trade practices, labor law violations, and Colorado Privacy Act violations. Before each use of the system to make or substantially factor in a consequential decision, deployers must provide consumers with detailed pre-decision disclosures identifying the system's trade name, developer, the personal aspects it evaluates, and how to access accommodation information. After an adverse consequential decision, deployers must provide a single notice within 30 days disclosing the main reasons, data categories and sources, information on exercising rights, and a copy of the notice. The appeal right is now limited to adverse decisions that are not competitive or time-limited and are adverse based on incorrect personal data or unlawful information. A new subsection (6.7) limits the most burdensome obligations — risk management programs, impact assessments, data correction rights, and appeal rights — to systems that are the principal basis in making a consequential decision (i.e., without meaningful human involvement). Tiered small-deployer exemptions phase in from 500 FTE to 100 FTE. A separate hiring-only exemption further phases down. The section expressly disclaims creating a private right of action.

Compliance actions 8 items

Deployers must implement and maintain an iterative risk management policy and program specifying the principles, processes, and personnel used to identify, document, and mitigate known or reasonably foreseeable risks of algorithmic discrimination, reviewed and updated regularly over the system's lifecycle. This obligation applies only to high-risk AI systems that are the principal basis of consequential decisions.

G-01.1

Deployers must complete an impact assessment for each high-risk AI system prior to first deployment (or January 1, 2027, whichever is later) and annually thereafter, covering risks of algorithmic discrimination, accessibility limitations, unfair trade practices, labor law violations, and Colorado Privacy Act violations, along with data categories, sources, outputs, and performance metrics. This obligation applies only to systems that are the principal basis of consequential decisions.

H-02.3

Deployers must conduct an annual review of each deployed high-risk AI system to verify it is not causing algorithmic discrimination.

H-02.8

Deployers must, before each use of a high-risk AI system to make or substantially factor in a consequential decision about a consumer, disclose the system's purpose, trade name, developer name, deployer contact information, a plain-language description including the roles of AI and human components, the personal aspects evaluated, the evaluation method, relevance to the decision, accommodation information, and instructions for accessing the deployer's public statement.

H-01.1

Deployers must, within 30 days after an adverse consequential decision, provide a single notice disclosing the main reasons for the decision (including the AI system's contribution and the categories and sources of data that adversely affected the output, including sensitive data), information on how to exercise correction and appeal rights, and a copy of the notice. Deployers must offer an opportunity to correct incorrect personal data. For non-competitive, non-time-limited adverse decisions based on incorrect data or unlawful information, deployers must provide an opportunity to appeal with human review if technically feasible. The correction and appeal rights apply only to systems that are the principal basis of the decision.

H-01.1

Deployers must not use a high-risk AI system to make or substantially factor in a consequential decision if they cannot provide accurate pre-decision and adverse-decision disclosures as required.

H-01

Deployers must publish and maintain on their website a clear and readily available statement summarizing the types of high-risk AI systems they deploy and how they manage associated risks.

G-02.4

Deployers must maintain all required documentation and records throughout deployment and for at least three years after final deployment, and must produce risk management policies, impact assessments, and records to the attorney general within 90 days of request.

R-02.2

C.R.S. § 6-1-1704

Disclosure of AI system to consumer

DeployerDeveloper

(1) 14 On and after January 1, 2027, except as provided in subsection (2) of this section, a deployer or other developer that deploys, offers, sells, leases, licenses, gives, or otherwise makes available an artificial intelligence system that is intended to interact with consumers shall disclose to each consumer who interacts with the artificial intelligence system that the consumer is interacting with an artificial intelligence system.

Section 4 amends the AI identity disclosure requirement. Beginning January 1, 2027, deployers or other developers that deploy, offer, sell, or otherwise make available an AI system intended to interact with consumers must disclose to each consumer that they are interacting with an artificial intelligence system. The amendment simplifies the obligation from "ensure the disclosure" to "disclose," placing the duty directly on the deployer or developer rather than requiring them to ensure a system-level disclosure mechanism.

Compliance actions 1 item

Deployers or developers must disclose to each consumer who interacts with an AI system intended to interact with consumers that they are interacting with an artificial intelligence system.

T-01.1

C.R.S. § 6-1-1705

Compliance with other legal obligations — exceptions and carve-outs

DeployerDeveloper

(1)(d.5), (1)(f), (1)(h), (1)(j), (1)(k) Nothing in this part 17 restricts a developer's, a deployer's, or other person's ability to: (d.5) Prosecute or defend legal claims during ongoing or imminently anticipated legal proceedings, including complying with the rules of procedure, rules of evidence, or other applicable rules or orders before a court, an administrative enforcement agency, or other legal tribunal of competent jurisdiction; (f) Except for uses of facial recognition technology otherwise prohibited by applicable law, prevent, detect, protect against, or respond to security incidents or illegal or tortious activity such as identity theft or fraud or investigate, report, or prosecute the persons responsible for that illegal or tortious activity; (h) Conduct research, testing, and development activities regarding an artificial intelligence system or model, other than testing conducted under real-world conditions, before the artificial intelligence system or model is used to make, or is used as a substantial factor in making, a consequential decision or is otherwise placed on the market, deployed, or put into service, as applicable; (j) Effectuate a product recall; or (k) Identify and repair technical errors that impair existing or intended functionality.

(6) Nothing in this part 17 applies to any artificial intelligence system to the extent that the artificial intelligence system: (a) Is acquired by or for the federal government or any federal agency or department, including the United States department of commerce, the United States department of defense, or the national aeronautics and space administration; (b) Is necessary to comply with applicable federal law; or (c) Has been specifically approved by a federal agency or department for use in making a consequential decision.

(10) 15 If a developer or deployer withholds information pursuant to a provision in this section for which disclosure would otherwise be required by this part 17, the developer or deployer shall notify the person that would otherwise have a right to receive the information, state the basis for withholding the information, cite the provision that authorizes the withholding of the information, and provide all information to which the basis for withholding does not apply. The notification must comply with the requirements of section 6-1-1703 (4)(c).

Section 5 revises the exceptions and carve-outs for the AI consumer protection framework. The bill narrows the legal-claims exception to ongoing or imminently anticipated proceedings (rather than the broader "investigate, establish, exercise, prepare for, or defend" language), broadens the security-incident exception to cover illegal or tortious activity, clarifies that compliance with Part 17 does not waive evidentiary privileges, and adds exceptions for product recalls and repair of technical errors. Several prior exceptions — federal-agency approval, federal research support, DoD contracts, and HIPAA-covered entities — are repealed, replaced in part by new language providing that Part 17 does not apply to AI systems acquired by the federal government, necessary to comply with federal law, or specifically approved by a federal agency for use in consequential decisions. A new withholding-notification obligation requires any developer or deployer that withholds information under this section to cite the authorizing provision and provide all non-withheld information.

Compliance actions 1 item

Developers or deployers that withhold information under a § 6-1-1705 exception must notify the intended recipient, state the basis for withholding, cite the authorizing provision, and provide all information to which the withholding basis does not apply.

G-02

C.R.S. § 6-1-1706

Enforcement by attorney general

(1)–(2) Notwithstanding section 6-1-103, the attorney general has exclusive authority to enforce this part 17 and may investigate and enforce violations of this part 17 beginning on January 1, 2027. Except as provided in subsection (3) of this section, each violation of the requirements established in this part 17 constitutes an unfair trade practice pursuant to section 6-1-105 (1)(hhhh).

(3) In any action commenced by the attorney general to enforce this part 17, it is an affirmative defense that the developer, deployer, or other person: (a) Discovered a curable violation of this part 17 as a result of: (III) An internal review process; and (a.5) Cured the violation described in subsection (3)(a) of this section within seven days after its discovery; (b) Was at all relevant times otherwise in compliance with this part 17 and: (III) Any risk management framework for artificial intelligence systems that the attorney general has designated and publicly disseminated; and (c) Demonstrates that the violation of this part 17 was inadvertent, affected fewer than one thousand consumers, and was not the result of negligence on the part of the developer, the deployer, or other person asserting the defense.

(5) Nothing in this part 17, including the enforcement authority granted to the attorney general under this section, preempts or otherwise affects any right, claim, remedy, presumption, or defense available at law or in equity. An affirmative defense established under this part 17 applies only to an enforcement action brought by the attorney general pursuant to this section and does not apply to any right, claim, remedy, presumption, or defense available at law or in equity.

Section 6 amends the enforcement provision. The attorney general retains exclusive enforcement authority and may investigate and enforce violations beginning January 1, 2027. Each violation constitutes an unfair trade practice under § 6-1-105(1)(hhhh). The affirmative defense is restructured: the defendant must show it (a) discovered a curable violation through an internal review process, (a.5) cured it within seven days, (b) was at all relevant times otherwise in compliance with Part 17 and any designated risk management framework, and (c) demonstrates the violation was inadvertent, affected fewer than 1,000 consumers, and was not the result of negligence. The defendant bears the burden of proof. The affirmative defense applies only to AG enforcement actions and does not affect any other rights, claims, or remedies at law or in equity.

C.R.S. § 6-1-1707

Rules — attorney general rulemaking authority

(1) The attorney general may adopt rules as necessary for the purpose of implementing and enforcing this part 17, including: (a) The documentation and requirements for developers pursuant to section 6-1-1702 (2); (b) The contents of and requirements for the notices and disclosures required by sections 6-1-1702 (3); 6-1-1703 (3) and (4); and 6-1-1704; (c) The content and requirements of the risk management policy and program required by section 6-1-1703 (2); (d) The content and requirements of the impact assessments required by section 6-1-1703; (e) The requirements for the affirmative defense set forth in section 6-1-1706 (3), including the process by which the attorney general will recognize any other nationally or internationally recognized risk management framework for artificial intelligence systems; and (f) Clarification of what constitutes a "consequential decision", as defined in section 6-1-1701 (3).

Section 7 updates the attorney general's rulemaking authority to align with the restructured obligations. The AG may adopt rules on developer documentation requirements, deployer notices, disclosures, risk management policies, impact assessments, the affirmative defense (including the process for recognizing risk management frameworks), and — newly added — clarification of what constitutes a consequential decision. References to the repealed rebuttable-presumption provisions are removed.

Section 8

Effective date

This act takes effect at 12:01 a.m. on the day following the expiration of the ninety-day period after final adjournment of the general assembly; except that, if a referendum petition is filed pursuant to section 1 (3) of article V of the state constitution against this act or an item, section, or part of this act within such period, then the act, item, section, or part will not take effect unless approved by the people at the general election to be held in November 2026 and, in such case, will take effect on the date of the official declaration of the vote thereon by the governor.

Section 8 provides that the act takes effect at 12:01 a.m. on the day following the expiration of the ninety-day period after final adjournment of the general assembly, subject to a referendum petition. If a referendum petition is filed, the act takes effect only if approved by voters at the November 2026 general election.

Transparency AI identity disclosure · periodic reminders

1 item

T-01.1

Beginning Jan 1, 2027 Deployers or developers must disclose to each consumer who interacts with an AI system intended to interact with consumers that they are interacting with an artificial intelligence system.

Human Oversight Decision review · appeal pathways

5 items

H-02.3

Beginning Jan 1, 2027 Deployers must complete an impact assessment for each high-risk AI system prior to first deployment (or January 1, 2027, whichever is later) and annually thereafter, covering risks of algorithmic discrimination, accessibility limitations, unfair trade practices, labor law violations, and Colorado Privacy Act violations, along with data categories, sources, outputs, and performance metrics. This obligation applies only to systems that are the principal basis of consequential decisions.

H-02.8

Beginning Jan 1, 2027 Deployers must conduct an annual review of each deployed high-risk AI system to verify it is not causing algorithmic discrimination.

H-01.1

Beginning May 1, 2026 Deployers must, before each use of a high-risk AI system to make or substantially factor in a consequential decision about a consumer, disclose the system's purpose, trade name, developer name, deployer contact information, a plain-language description including the roles of AI and human components, the personal aspects evaluated, the evaluation method, relevance to the decision, accommodation information, and instructions for accessing the deployer's public statement.

H-01.1

Beginning May 1, 2026 Deployers must, within 30 days after an adverse consequential decision, provide a single notice disclosing the main reasons for the decision (including the AI system's contribution and the categories and sources of data that adversely affected the output, including sensitive data), information on how to exercise correction and appeal rights, and a copy of the notice. Deployers must offer an opportunity to correct incorrect personal data. For non-competitive, non-time-limited adverse decisions based on incorrect data or unlawful information, deployers must provide an opportunity to appeal with human review if technically feasible. The correction and appeal rights apply only to systems that are the principal basis of the decision.

H-01

Deployers must not use a high-risk AI system to make or substantially factor in a consequential decision if they cannot provide accurate pre-decision and adverse-decision disclosures as required.

Governance Risk management · accountability programs

7 items

G-02.1

Beginning Jan 1, 2027 Developers must make available to each deployer or other developer a general statement describing intended uses and known harmful uses of the high-risk AI system, along with documentation describing data governance measures, intended inputs, and outputs.

G-02.1

Beginning Jan 1, 2027 Developers must make available to deployers, to the extent feasible, the documentation and information (such as model cards, dataset cards, or impact assessments) necessary for deployers to complete an impact assessment under § 6-1-1703(3).

G-02.4

Beginning Jan 1, 2027 Developers must publish and maintain on their website or in a public use case inventory a statement summarizing the types of high-risk AI systems they offer and how they manage known or reasonably foreseeable risks of algorithmic discrimination.

G-02

Beginning Jan 1, 2027 Developers that withhold information from a required disclosure based on trade secret, legal protection, or security risk must notify the intended recipient, state the basis for withholding, and provide all information to which the basis does not apply.

G-01.1

Beginning Jan 1, 2027 Deployers must implement and maintain an iterative risk management policy and program specifying the principles, processes, and personnel used to identify, document, and mitigate known or reasonably foreseeable risks of algorithmic discrimination, reviewed and updated regularly over the system's lifecycle. This obligation applies only to high-risk AI systems that are the principal basis of consequential decisions.

G-02.4

Beginning Jan 1, 2027 Deployers must publish and maintain on their website a clear and readily available statement summarizing the types of high-risk AI systems they deploy and how they manage associated risks.

G-02

Regulatory Reporting · disclosure · audit

2 items

R-02.2

Beginning Jan 1, 2027 Developers must maintain all required documentation and records throughout the distribution period and for at least three years after final distribution, and must produce them to the attorney general within 90 days of request.

R-02.2

Beginning Jan 1, 2027 Deployers must maintain all required documentation and records throughout deployment and for at least three years after final deployment, and must produce risk management policies, impact assessments, and records to the attorney general within 90 days of request.

Passage Likelihood

Failed

Status Failed

Final action Senate Committee on Business, Labor, & Technology Postpone Indefinitely

Legislative History

2025-04-28 Introduced In Senate - Assigned to Business, Labor, & Technology

2025-05-05 Senate Committee on Business, Labor, & Technology Postpone Indefinitely

Sources

Full Text ↗

Entry Last Reviewed

2026-05-19

AI generated