The regulatory landscape for Artificial Intelligence (AI) in Colorado may be shifting. Following the enactment of the Colorado AI Act (SB24-205), the AI Policy Working Group, comprised of technology industry representatives, consumer advocates, and business groups, has introduced a proposed bill that would modify the existing law currently scheduled to take effect on June 30, 2026.
The proposal introduces a more targeted framework intended to narrow and clarify employer compliance obligations, making them more manageable and administratively feasible. If enacted, it would delay the effective date to January 1, 2027.
How the Proposed Bill Differs from SB24-205
The primary objective of the new proposal is to reduce the technical and operational burdens placed on deployers of AI technologies. While the prior law imposed broad risk management and mandatory state reporting, the new framework focuses on transparency and consumer rights regarding “consequential decisions” such as hiring, terminations, and compensation.
Key differences between the two versions include:
- Elimination of Risk Management Programs. SB24-205 required employers to implement and maintain a comprehensive, formal risk management policy. The proposed bill removes this internal mandate, focusing instead on clear disclosures.
- Removal of Mandatory Impact Assessments. The previous requirement to perform and document annual, detailed impact assessments for high-risk systems is absent in the new proposal.
- Revised Duty of Care. Under SB24-205, employers had a proactive duty to report any discovery of “algorithmic discrimination” to the Attorney General within 90 days. The proposed bill shifts the focus toward providing individuals with notice and the right to human review.
- Shared Liability Framework. The proposed bill introduces a more defined approach to liability when AI-related issues arise. Unlike the prior law, which left responsibility more ambiguous, this framework allocates liability based on fault. Developers may be held accountable where harm results from the system’s intended use, while employers (as “deployers”) are responsible for independent decisions, such as using the AI in ways not authorized by the developer. Notably, the bill would render void any contractual provisions that attempt to shift one party’s liability to another.
- Specific Recordkeeping. The proposal introduces a clear three-year retention period for system version identifiers and change logs, providing a more defined compliance target than the broader language in the prior law.
Key Requirements Under the New Framework
If enacted, the proposed bill would establish specific obligations for employers using automated decision making technology that “materially influences” an employment decision.
- Notice and Transparency. Employers must provide conspicuous “point-of-interaction” notices to individuals, such as job applicants, informing them that automated technology will be used. If a decision results in an adverse outcome (e.g., a candidate is not selected), the employer must provide a plain-language description of the technology’s role in that outcome.
- Right to Human Review. A central pillar of the proposal is the right to a “meaningful human review.” Employers must establish a process where a person with the authority to change the outcome can review and reconsider adverse decisions influenced by AI.
- Correction of Personal Data. Individuals have the right to request correction of any factually inaccurate personal data used by the technology in reaching a decision.
Recommended Next Steps for Employers
Although this is currently a proposed framework, employers should begin evaluating their technology to ensure a smooth transition should the bill become law.
- Inventory Automated Systems: Identify all software used in recruitment, performance management, or compensation that meets the definition of “materially influencing” a decision.
- Evaluate Vendor Documentation: Review contracts with software developers to ensure they provide the necessary version identifiers and change logs required for the three-year recordkeeping mandate.
- Draft Disclosure Templates: Begin creating plain-language notice templates for job applications and adverse action letters to ensure transparency.
- Define Human Review Protocols: Identify which internal stakeholders will be responsible for conducting the “meaningful human review” of contested decisions.
- Monitor Legislative Status: Stay informed on the bill’s progress.
HR Works provides industry expertise and guidance to HR practitioners navigating the evolving legislation of artificial intelligence. Our governance tools include inventory tracking, disclosure templates, vendor vetting protocols, and legislative updates, within a suite of other resources designed to outfit HR with the information needed to keep up with this rapidly developing field.
Contact us if you would like more information.
