Table of Contents
State Tech Pulse Policy Brief: AI Oversight vs. Innovation
Background: In the last five years, state legislators have been in the driver's seat on discussions and deliberations around how artificial intelligence (AI) should be regulated. Stakeholders have engaged in complicated negotiations around the increasing tension between regulation and innovation. This policy brief explores the 5 W’s (Who, What, Where, When, Why?) regarding the first state (Colorado) that enacted the US’s first comprehensive AI legislation.
Public Policy Problem: The lifecycle of an AI system includes its design, development, deployment, and use. Risks can develop across multiple points within the lifecycle, influenced and impacted by developers, deployers, end users, and the public. There is a debate over how and when AI should be regulated by the government. Leaders have been looking at AI through an intersectional lens and critically evaluating whether certain measures should be in place for all AI applications or whether certain applications should be focused on those that present more risk.
Proposed Policy Solutions:
Evidence from proposals at the state, federal, and international levels shows that leaders at all levels are grappling with many legal and societal impacts the technology can have. Strategies across the states vary. Themes of these considerations span a wide spectrum, from privacy and data protection to questions of liability and concerns about bias and discrimination, among other factors.
Here are some questions leaders are contemplating:
Do current laws on the books need to be modified to include emerging technology applications, or can already established laws apply to these new uses?
If existing laws already apply, do states need to add definitions of what artificial intelligence is?
Will attempting to create new definitions for an evolving technology work?
Should legislators approach AI regulation in a “comprehensive” manner or narrow new laws to apply to specific sectors and applications?
Who? In Colorado, over the past few years, stakeholders have engaged with state leaders on AI regulation. Feedback has been presented in formal and informal ways. In 2024, after Colorado passed the most comprehensive AI law in the nation, a task force was quickly stood up to address differences among stakeholders.
Negotiations and compromises were attempted, with stakeholders' views ranging from the role of government in the deployment and oversight of this technology.
The Colorado business community has been engaged in providing feedback to leaders on AI regulations.
This month, a bipartisan coalition of Colorado technology and business leaders penned a letter to the Governor and other state leaders with concern about the technology economy and business climate. The letter calls for leaders to take multiple actions, including conducting an assessment of regulatory and legislative factors contributing to Colorado “losing ground to competing states.”
Stakeholders, spanning civil society, labor advocates, and others, presented their feedback to legislators and task force members. A coalition of Colorado- and nationally based organizations summarized their feedback on Colorado’s new law. Feedback includes what these stakeholders like from the law, including the comprehensive nature of the law, impact assessments, and the right to explain the reasons behind an AI decision. Areas where advocates sought changes included strengthening impact assessment requirements and expanding the law’s transparency provisions.
What? Colorado enacted SB 205, a comprehensive AI law that applies to high-risk systems and requires developers and deployers to avoid algorithmic discrimination in consequential decisions in areas such as education, employment, financial or lending services, essential government services, healthcare services, housing, insurance, or legal services. The new law adds new requirements developers and deployers must meet, such as disclosing known inappropriate uses of the technology, the data used to train the system, implementing a risk management policy, and conducting an impact assessment. Following the passage of this bill, Colorado’s legislature created a task force aimed at working through differences and finding consensus to make changes to this law.
Where? States are looking to Colorado to see if the most comprehensive US law will be upheld.
When? The timeline…
April 2021: The European Commission released a draft to regulate AI in the European Union. The EU AI Act is a risk-based approach that includes obligations related to impact assessments. Colorado’s comprehensive AI law gleaned inspiration from these regulations.
December 2023: The European Parliament and Council reached agreement on the legislative language to move forward.
May 2024: The European Council adopted the EU AI Act.
May 2024: The Colorado Artificial Intelligence Act was enacted by the Colorado General Assembly and signed by the Governor.
June 2024: Colorado’s Artificial Intelligence Impact Task Force was created and tasked with coming up with recommendations by February 2025. The scope of what the task force aimed to study ranged from defining AI to developing recommendations for how the state can effectively govern AI systems and automated decision-making systems.
August 2025: Colorado’s Governor called a special session to address proposed changes to Colorado’s AI Act. The only change made during this time was pushing back the implementation of the law from Feb. 1, 2026, to June 30, 2026.
February 2025: The Colorado Artificial Intelligence Impact Task Force submitted a report containing its recommendations. According to the report, the task force suggested that further negotiations are needed and identified areas of consensus and major disagreements. Since the members could not reach a consensus, the report found that the task force was unable to make substantive recommendations.
March 2025: The Virginia General Assembly was the second state to pass a comprehensive AI law (HB25- 2094), similar to Colorado’s, which the Governor ended up vetoing.
October 2025: Colorado’s Governor created an “Artificial Intelligence policy working group” to develop a policy framework that the legislature can consider and advance during the 2026 legislative session. In a letter to participants, the Governor urged them to look beyond Colorado’s AI act to other solutions to propose changes to existing law. This group was handpicked by the Governor, a lot smaller than the task force the legislature created, and only included a few participants from the task force.
November 2025: The European Commission released proposals to make changes to the EU AI Act and the General Data Protection Regulation.
December 2025: President Trump signed an executive order urging Congress to work with him on a “minimally burdensome national standard” that would supersede conflicting and burdensome state laws. According to the document, states with onerous laws may not receive federal funding tied to increased access and deployment of broadband.
March 2026: President Trump released a “National Policy Framework for Artificial Intelligence,” which specifies the preference for a federal strategy versus state-led AI regulations. The report also focuses on six other areas that the federal standard should address:
Youth-focused protections
AI infrastructure and small business support
Intellectual property
Censorship and free speech
Innovation
Workforce preparation
March 2026: The Governor-led Colorado AI Policy Work Group released a proposal to revise the Colorado AI Act, which received unanimous support from the working group.
Why? The Governor-led working group's proposal to change Colorado’s AI law includes major changes from the current law.
The new version removes the definitions of "algorithmic discrimination” and “artificial intelligence system.” In its place, the group suggests adding the term “automated decision-making technology,” which is a much narrower definition and aligns with terms in state-led data privacy laws.
The new version retains the definition of "consequential decisions,” but expands on what is not included in a "consequential decision.” The new proposal expands on use cases that the new proposal defines does not fall into this definition. Some new additions of what would not qualify are routine scheduling, classroom personalization, and workflow management. Additionally, the proposal removes AI decisions related to access to legal services.
Current law requires reporting of algorithmic discrimination, conducting AI impact assessments, and implementing a risk management policy, among other obligations. The proposed rewrite from the Governor-led group focuses on transparency, recordkeeping, and consumer rights, while aligning with laws already in Colorado law. For example, the proposed language is modeled after state-led privacy policies, focusing on notice, disclosures, and reviews.
Enforcement and further rulemaking to clarify new regulations remain with the Attorney General. The new proposal expands what the Attorney General must clarify through rulemaking, such as sector-specific guidance or illustrative examples tailored to different consequential decision domains, including employment, housing, lending, insurance, education, health care, and government services. None of the versions includes a private right of action.
If no action is taken, Colorado’s AI law, in its current version, will become effective in June 2026. The new proposal includes an adjusted timeline, with a suggested implementation date of January 2027.
Colorado’s 2026 legislative session is scheduled to end on May 13th, with the proposed language still awaiting introduction by a legislator. Other AI legislation is moving through the legislative process focused on more narrow use cases, such as AI and psychotherapy, AI chatbots,use of AI in health care, and surveillance pricing.
Do you have leads, tips, corrections, feedback or resources you would like to share? Send your advice to [email protected].
Disclosure: This is a human-written and driven publication. As a small business owner and mighty team of one, I use AI tools to optimize my small business operations as a part of my admin tech stack. Regarding this publication, AI is mainly used to help with catchy titles, as a thesaurus when writing and a partner when creating cartoons. (Thanks, Canva, and not an ad!) As a secret doodler, I add my human touch using my digital pad and pen. I also use Grammarly, with AI built in, to help with copy editing/grammar (again, mighty team of one!) Thanks for reading. 😊