Blab AI
Colorado Prepares Rulemaking on AI-Driven Employment Tools Ahead of 2027 Compliance Deadline
Colorado’s AI watchdog is tightening the reins on hiring robots. In August 2026, the state Attorney General’s Office filed draft rules that will clarify how the new Automated Decision‑Making Technology (ADMT) law applies to the AI tools employers use to hire, promote and fire workers. The rules will be debated at a public hearing on October 26 2026 and are expected to shape the compliance obligations for Colorado employers by the law’s effective date of January 1 2027.
The proposed rules, titled Automated Decision‑Making Technology and Conversational Artificial Intelligence Service Rules, build on Senate Bill 26‑189 (the ADMT Act) and House Bill 26‑1263 (the Chatbot Safety Act). Both bills were signed by Governor Jared Polis in May 2026 and replace the state’s 2024 AI law. The ADMT Act introduces a notice‑and‑transparency framework that requires deployers of AI tools to inform affected individuals, provide explanations for adverse outcomes, and allow for human review.
A central focus of the draft is the definition of “materially influence” – the threshold that triggers the ADMT Act’s obligations. The law defines a material influence as an ADMT output that is a non‑de‑minis factor affecting a consequential decision, such as hiring or firing. The Attorney General’s office is soliciting input on how to interpret “non‑de‑minis.” Two competing standards are under consideration:
1. Lower Threshold – A factor is de‑minis only if it has a trivial or incidental impact, regardless of other factors. Under this standard, a rebuttable presumption of material influence arises whenever the output pertains to the individual, is reviewed by the decision‑maker, and aligns with the outcome. Employers could rebut by showing independent judgment or reliance on other documented factors. 2. Higher Threshold – A factor is de‑minis if it is not a substantial factor in the decision, meaning other independent factors play a significantly larger role. The presumption of material influence remains the same, but rebuttal requires evidence of an independent review of substantial information and the decision‑maker’s authority and subject‑matter understanding.
The choice between these standards will determine how many AI‑based hiring tools fall within the law’s scope. A lower threshold would likely bring a large portion of employment‑related ADMTs under regulatory oversight.
The draft rules also address multiparty arrangements that are common in the employment context. The ADMT Act does not explicitly cover situations where an employer relies on a third‑party vendor’s AI tool. The Attorney General’s office is asking whether rules should:
Recognize that a deployer may run an ADMT through a vendor; Require deployers to remain responsible for all obligations regardless of vendor involvement; Mandate vendor assistance in responding to consumer data and correction requests; Clarify responsibilities in staffing‑agency scenarios where the agency operates the ADMT but the employer uses its output; * Model vendor obligations after processor duties under the Colorado Privacy Act.
These questions are critical because they will shape contractual arrangements, cost structures, and risk allocation for employers who outsource AI decision‑making.
Finally, the rules seek to clarify developer responsibilities. The ADMT Act requires developers to provide certain information to deployers, but the draft rules ask whether additional disclosure is needed on testing procedures, training data categories, and known limitations. More detailed disclosures could help employers evaluate and compare AI tools before procurement.
The public hearing on October 26 2026 will allow stakeholders to comment on the draft rules. Written comments are due by September 4 2026, and a second draft will circulate by September 23 2026. Any comments received on or before the hearing’s closing time will be considered in the final rules. If the rulemaking process extends beyond the hearing, the Attorney General will accept comments until the last day of formal rulemaking.
Colorado’s move reflects a broader trend of state‑level AI regulation that targets employment‑related decision‑making. The proposed rules will provide a clearer framework for employers operating in Colorado, helping them navigate the upcoming compliance deadline and avoid potential legal exposure.
In summary, Colorado’s Attorney General is preparing to finalize rules that will define when AI tools used in employment decisions trigger the ADMT Act’s notice and transparency requirements. The rules will address the material‑influence threshold, responsibilities in vendor‑based arrangements, and developer disclosure obligations. Stakeholders are encouraged to submit comments before the October 26 hearing to influence the final regulatory framework that will take effect on January 1 2027.
The proposed rules, titled Automated Decision‑Making Technology and Conversational Artificial Intelligence Service Rules, build on Senate Bill 26‑189 (the ADMT Act) and House Bill 26‑1263 (the Chatbot Safety Act). Both bills were signed by Governor Jared Polis in May 2026 and replace the state’s 2024 AI law. The ADMT Act introduces a notice‑and‑transparency framework that requires deployers of AI tools to inform affected individuals, provide explanations for adverse outcomes, and allow for human review.
A central focus of the draft is the definition of “materially influence” – the threshold that triggers the ADMT Act’s obligations. The law defines a material influence as an ADMT output that is a non‑de‑minis factor affecting a consequential decision, such as hiring or firing. The Attorney General’s office is soliciting input on how to interpret “non‑de‑minis.” Two competing standards are under consideration:
1. Lower Threshold – A factor is de‑minis only if it has a trivial or incidental impact, regardless of other factors. Under this standard, a rebuttable presumption of material influence arises whenever the output pertains to the individual, is reviewed by the decision‑maker, and aligns with the outcome. Employers could rebut by showing independent judgment or reliance on other documented factors. 2. Higher Threshold – A factor is de‑minis if it is not a substantial factor in the decision, meaning other independent factors play a significantly larger role. The presumption of material influence remains the same, but rebuttal requires evidence of an independent review of substantial information and the decision‑maker’s authority and subject‑matter understanding.
The choice between these standards will determine how many AI‑based hiring tools fall within the law’s scope. A lower threshold would likely bring a large portion of employment‑related ADMTs under regulatory oversight.
The draft rules also address multiparty arrangements that are common in the employment context. The ADMT Act does not explicitly cover situations where an employer relies on a third‑party vendor’s AI tool. The Attorney General’s office is asking whether rules should:
Recognize that a deployer may run an ADMT through a vendor; Require deployers to remain responsible for all obligations regardless of vendor involvement; Mandate vendor assistance in responding to consumer data and correction requests; Clarify responsibilities in staffing‑agency scenarios where the agency operates the ADMT but the employer uses its output; * Model vendor obligations after processor duties under the Colorado Privacy Act.
These questions are critical because they will shape contractual arrangements, cost structures, and risk allocation for employers who outsource AI decision‑making.
Finally, the rules seek to clarify developer responsibilities. The ADMT Act requires developers to provide certain information to deployers, but the draft rules ask whether additional disclosure is needed on testing procedures, training data categories, and known limitations. More detailed disclosures could help employers evaluate and compare AI tools before procurement.
The public hearing on October 26 2026 will allow stakeholders to comment on the draft rules. Written comments are due by September 4 2026, and a second draft will circulate by September 23 2026. Any comments received on or before the hearing’s closing time will be considered in the final rules. If the rulemaking process extends beyond the hearing, the Attorney General will accept comments until the last day of formal rulemaking.
Colorado’s move reflects a broader trend of state‑level AI regulation that targets employment‑related decision‑making. The proposed rules will provide a clearer framework for employers operating in Colorado, helping them navigate the upcoming compliance deadline and avoid potential legal exposure.
In summary, Colorado’s Attorney General is preparing to finalize rules that will define when AI tools used in employment decisions trigger the ADMT Act’s notice and transparency requirements. The rules will address the material‑influence threshold, responsibilities in vendor‑based arrangements, and developer disclosure obligations. Stakeholders are encouraged to submit comments before the October 26 hearing to influence the final regulatory framework that will take effect on January 1 2027.
