Asep Darmawan
Denver, CO – Colorado is once again taking a prominent, albeit revised, stance in the burgeoning field of artificial intelligence regulation. Following significant debate and legal challenges to its initial foray into AI governance, the state has enacted a new law, Senate Bill 26-189, designed to establish a framework for the transparent use of automated decision-making technologies (ADMTs). This legislative pivot moves away from the contentious concept of "algorithmic discrimination" that defined its predecessor, Senate Bill 24-205, opting instead for a broader definition of regulated technologies and enhanced individual rights concerning their deployment.
The genesis of SB 26-189 lies in the tumultuous reception of SB 24-205, a bill that, despite its passage, sparked considerable concern among industry leaders, including tech giants like Elon Musk’s xAI. The original legislation aimed to prevent "algorithmic discrimination" in employment contexts, a term that proved to be a significant sticking point. Critics argued that the vagueness of "algorithmic discrimination" posed a threat to innovation and could lead to unintended consequences for businesses operating within the state. This sentiment was echoed by state officials, including Governor Jared Polis, who acknowledged a desire to revisit the initial law. The tension culminated in a lawsuit filed by xAI against Colorado’s attorney general, seeking to block the enforcement of SB 24-205, citing its alleged unconstitutionality due to the ill-defined nature of its core prohibition.
SB 26-189 represents a deliberate effort to address these criticisms while still providing a mechanism for oversight and individual recourse. The new law completely removes references to "algorithmic discrimination" and introduces a more encompassing definition of the technologies it seeks to regulate. This shift signals a pragmatic approach, aiming to strike a balance between fostering technological advancement and safeguarding individuals from potentially unfair or opaque automated processes.
A Chronology of Colorado’s AI Legislative Journey
The path to SB 26-189 has been a dynamic one, marked by legislative action, public discourse, and legal maneuvering.
- Early 2024: Colorado lawmakers introduce Senate Bill 24-205, aiming to regulate "high-risk artificial intelligence systems" and prohibit algorithmic discrimination in employment. The bill garners significant attention, both positive and negative, from various stakeholders.
- Spring 2024: SB 24-205 passes the Colorado legislature. However, concerns about its broad scope and the ambiguity of key terms begin to surface. State officials, including Governor Polis, express a willingness to re-evaluate the law.
- Late Spring/Early Summer 2024: Elon Musk’s xAI files a lawsuit against Colorado’s Attorney General, challenging the constitutionality of SB 24-205. The lawsuit specifically targets the term "algorithmic discrimination" as being unconstitutionally vague. This legal action amplifies the debate surrounding the law’s enforceability and its potential impact on the AI industry.
- Summer 2024: In response to the ongoing debates and legal challenges, Colorado lawmakers introduce Senate Bill 26-189. This new legislation is crafted to replace SB 24-205, incorporating feedback and addressing the concerns raised by industry and legal experts.
- Late Summer 2024: SB 26-189 passes the Colorado legislature. The bill focuses on transparency and establishes a new definition of "automated decision-making technology" (ADMT), moving away from the problematic "algorithmic discrimination" clause.
- [Date of Article Publication]: Governor Jared Polis signs SB 26-189 into law. The effective date for the law is set for 2027, allowing time for the Attorney General to develop necessary regulations.
This timeline highlights Colorado’s proactive, yet iterative, approach to AI regulation. The state’s willingness to amend its legislative course in response to feedback and legal challenges underscores a commitment to finding a workable solution in a rapidly evolving technological landscape.
Defining the Scope: From "High Risk" to "Automated Decision-Making Technology"
A significant departure in SB 26-189 is its redefinition of the technologies under its purview. While SB 24-205 focused on "high risk artificial intelligence systems," the new law adopts a broader umbrella term: "automated decision-making technology," or ADMT.
This expanded definition encompasses "any tool that processes personal data and uses computation to generate output, including ‘predictions, recommendations, classifications, rankings, scores, or other information used to make guide, or assist a decision, judgment, or determination’ about an individual." This comprehensive approach aims to capture a wider array of automated systems that can impact individuals’ lives, moving beyond a narrow focus on explicitly labeled "AI systems."
However, SB 26-189 also carves out specific exclusions, aiming to prevent overregulation of less impactful technologies. Generally, tools are excluded if they are not used to make a "consequential decision." The law clarifies that "low-stakes or routine decisions" that do not "materially influence eligibility for, selection for, denial of, compensation for or access to employment and similarly covered opportunities" are not considered consequential.
Furthermore, the legislation explicitly states that it does not cover ADMTs used solely for summarizing, organizing, or presenting information for human review, provided they do not produce a score, ranking, recommendation, classification, prediction, or other inference that materially influences an outcome or decision. This distinction is crucial, ensuring that tools that merely facilitate human analysis, rather than automate decision-making with significant consequences, remain outside the law’s purview.
Empowering Individuals: Rights and Recourse Under the New Law
SB 26-189 introduces a set of rights for individuals who experience adverse outcomes from consequential decisions made by an ADMT. These provisions are designed to provide a degree of transparency and accountability.
Under the new law, individuals affected by such decisions may:
- Request Information: They can request instructions on how to obtain their personal data that was used in the decision-making process.
- Seek Correction: Individuals have the right to request the correction of factually inaccurate or materially inaccurate personal data that influenced the decision.
- Demand Human Review: Crucially, the law grants individuals the formal right to request an opportunity for "meaningful human review and reconsideration" of the decision made by the ADMT. This provision aims to ensure that automated decisions are not final and that human judgment can be applied when necessary.
These individual rights represent a significant step towards ensuring that automated systems are used responsibly and that individuals have a mechanism to challenge outcomes they believe to be unfair or based on flawed data.
Regulatory Framework and Enforcement
The implementation and enforcement of SB 26-189 will be overseen by Colorado’s Attorney General. The law mandates that the Attorney General will issue regulations to clarify and implement its requirements on or before the law’s effective date of 2027. This preparatory period is essential for developing clear guidelines and ensuring a smooth transition for businesses.
In the event of a violation, the law outlines a structured enforcement process. If the Attorney General determines that an organization has violated the law, the organization will be granted a 60-day period to "cure" the violation, if possible. This cure period is intended to allow businesses to rectify issues before facing legal action. However, the Attorney General retains the discretion to forgo the cure period requirement if the organization is found to have knowingly or repeatedly violated the law, indicating a tiered approach to enforcement that prioritizes intentional or persistent non-compliance.
Official Responses and the Path Forward
The passage of SB 26-189 has been met with positive reactions from the Colorado governor’s office and other stakeholders. Governor Jared Polis, in a press release, lauded the new law as a product of "collaborative effort between consumers and various other stakeholders." He emphasized its potential to preserve Colorado’s standing in the dynamic AI industry.
"This is a big step in the right direction for Colorado, and a model for the rest of the country," Governor Polis stated. "Replacing the old law that hasn’t taken effect yet will boost Colorado innovation and entrepreneurship." This statement reflects a belief that the revised legislation strikes a more effective balance, fostering innovation while still providing necessary safeguards.
The news of SB 26-189’s passage comes shortly after Colorado’s legislature enacted another significant bill, prohibiting employers from using AI or similar tools to set individualized wages, thereby preventing potential wage discrimination. This demonstrates a broader legislative focus on the ethical and equitable use of AI in the workplace.
Legal and industry analysts are advising affected employers to take proactive steps in response to the new law. Attorneys from Fisher Phillips, in an analysis, recommend that employers thoroughly map their AI tools and engage in discussions with their vendors to understand how their systems align with the new requirements. They also highlight the law’s three-year recordkeeping requirement for relevant compliance documents and advise businesses to remain vigilant for any potential legal challenges that may arise concerning the revised legislation.
Colorado’s journey in AI regulation is a compelling case study in legislative adaptation. By learning from the challenges posed by its initial attempt, the state has crafted a new framework that prioritizes transparency and individual recourse while seeking to avoid stifling the innovation that drives the burgeoning AI sector. The success of SB 26-189 will ultimately be measured by its ability to achieve this delicate balance and serve as a blueprint for other jurisdictions navigating the complex landscape of artificial intelligence.
Azzam Bilal Chamdy
Lina Irawan
Sagoh
Suro Senen
Lina Hope
Layla Zulfa
Raul Delapena Setiawan
Dwi Wanna