America’s most ambitious attempt to combat algorithmic bias was supposed to reshape AI deployment across the country. Instead, within weeks of its June 30, 2026 effective date, Colorado’s Senate Bill 24-205 — the Colorado Artificial Intelligence Act — found itself frozen by a federal court, challenged simultaneously by Elon Musk’s xAI startup and the United States Department of Justice. The episode is the most significant test yet of whether states can regulate AI independently, and the answer so far is: not easily.
The Law That Started It All
Colorado’s SB 24-205 was signed in 2024 and quickly earned the title of the most comprehensive AI anti-discrimination statute in the United States. Originally scheduled to take effect on February 1, 2026, the legislature pushed it back to June 30, 2026 to give industry more time to prepare.
The law targeted what regulators call “high-risk artificial intelligence systems” — AI tools that inform consequential decisions in areas such as education, employment, financial services, government services, healthcare, housing, insurance, and legal proceedings. Any AI system affecting Colorado residents fell within its scope, regardless of where the developer was headquartered.
Under SB 24-205, developers and deployers of high-risk AI systems had to:
- Exercise reasonable care to prevent algorithmic discrimination
- Conduct impact assessments with specific prescribed content and methodology
- Disclose to individuals when they are interacting with or being evaluated by an AI system
- Publish public disclosures on their websites explaining how high-risk systems are used
- Conduct annual reviews and maintain risk management programs
The law also included an exemption permitting AI systems explicitly designed to increase diversity or redress historical discrimination — a carveout that would become a central target for challengers.
For the AI industry, the compliance burden was daunting. For consumer advocates, it represented a meaningful floor of accountability. For two years, both sides watched the clock tick toward June 30.
xAI Pulls the Trigger
On April 9, 2026, xAI — the AI company founded by Elon Musk and creator of the Grok large language model — filed a federal lawsuit in Colorado challenging SB 24-205 on constitutional grounds. The complaint laid out four distinct legal theories:
First Amendment violations. xAI argued that designing AI models is a form of protected speech and that the law’s mandatory disclosure requirements amount to compelled speech — forcing the company to say things it does not wish to say, in violation of the First Amendment.
Commerce Clause overreach. The company contended the law improperly extends its reach to regulate out-of-state activity. Because compliance would require xAI to alter its AI products on a nationwide basis — not just for Colorado users — the law effectively imposes Colorado’s regulatory preferences on the entire country, which xAI argued violates the dormant Commerce Clause.
Due process vagueness. xAI targeted specific statutory language, arguing that undefined terms like “perceived” characteristics and “differential treatment” are so vague they invite arbitrary enforcement and fail to give regulated parties fair notice of what is prohibited.
Equal Protection discrimination. The company challenged the diversity carveout as itself unconstitutionally discriminatory — authorizing disparate treatment based on race and sex without a sufficient constitutional justification.
The lawsuit immediately put the law on trial before it ever took effect, and it attracted a co-plaintiff with far more institutional weight: the federal government itself.
The DOJ Enters — A Historic First
On April 24, 2026, the United States Department of Justice filed a Complaint in Intervention, joining the lawsuit against Colorado’s AI law. This marked the first time in U.S. history that the federal government formally intervened in litigation challenging a state-level AI regulation.
The DOJ’s intervention did not simply echo xAI’s arguments. It focused narrowly but powerfully on Equal Protection Clause violations, filing a complaint with two specific counts:
Count 1 — Compelled Discrimination. The DOJ alleged that SB 24-205’s disparate-impact liability framework compels AI developers to engage in race- and sex-conscious engineering of their models. To avoid liability under the statistical disparity provisions, developers must affirmatively engineer demographic outcomes — which the DOJ characterized as intentional discrimination against other groups in competitive contexts.
Count 2 — Authorized Discrimination. The DOJ directly attacked the diversity carveout, arguing it “authorizes intentional differential treatment based on protected classes without the constitutionally required justification.” In the DOJ’s reading, Colorado explicitly permits AI systems to treat people differently based on race or sex, and that permission is unconstitutional on its face.
The intervention is legally significant because it transforms the litigation from a private tech company grievance into a federal constitutional challenge backed by the sovereign power of the United States. Courts give considerable weight to government parties in constitutional litigation.
The DOJ’s action also fits within a broader pattern: in December 2025, President Trump signed an Executive Order establishing an “AI Litigation Task Force” explicitly tasked with identifying and challenging state AI laws deemed inconsistent with federal innovation priorities. Colorado’s law was an obvious first target.
The Enforcement Stay
Three days after the DOJ moved to intervene — on April 27, 2026 — a federal magistrate judge granted a joint motion to suspend enforcement of SB 24-205. The court imposed two conditions for lifting the stay:
- The conclusion of the 2026 Colorado General Assembly legislative session and any associated rulemaking process
- A decision on xAI’s pending motion for a preliminary injunction
In practical terms, this means SB 24-205 cannot be enforced even if it reaches its nominal June 30 effective date. The law is on the books but effectively inert — companies are not required to comply while the stay is in place.
Advertisement
Colorado Blinks: SB 26-189
Colorado lawmakers did not wait for the courts to resolve the constitutional questions. Moving with unusual speed, the legislature introduced SB 26-189 as a replacement framework. Both chambers passed it by May 9, 2026 — barely three weeks after the DOJ’s intervention.
SB 26-189 would repeal the original SB 24-205 framework entirely and replace its “high-risk AI system” architecture with a narrower concept: “automated decision-making technology” (ADMT). The new bill significantly scales back the original law’s ambition:
- The broad impact assessment and risk management requirements are stripped out
- The new framework focuses primarily on transparency obligations beginning January 1, 2027
- Companies must provide notice requirements to individuals affected by ADMT
- They must maintain three-year recordkeeping for automated decisions
- They must provide plain language descriptions of adverse automated decisions within 30 days
- Individuals gain rights to correction and human review of automated decisions
The shift from “high-risk AI system” to “automated decision-making technology” is more than a terminological swap — it reflects a fundamentally different theory of regulation. Rather than requiring pre-deployment impact assessments and ongoing risk management across a broad AI ecosystem, the new framework focuses on after-the-fact transparency and individual rights when something goes wrong.
Critics argue SB 26-189 guts the original law’s preventive architecture. Supporters say it creates a more defensible, narrower framework that can actually survive constitutional challenge.
Why This Matters Beyond Colorado
The Colorado episode has immediate implications for AI governance well beyond the state’s borders. At least 45 states have introduced some form of AI regulation in 2025-2026 legislative sessions, with many modeling their proposals on Colorado’s original framework. The lawsuit and DOJ intervention send a clear message: ambitious state AI legislation faces a federal veto threat.
Several dynamics make this particularly consequential:
The DOJ’s AI Litigation Task Force is active and looking for targets. Colorado was first, but it is unlikely to be last. Any state law that includes disparate-impact liability, diversity carveouts, or broad extraterritorial reach now faces a credible threat of federal challenge.
Constitutional arguments against state AI laws are well-developed and increasingly battle-tested. xAI’s lawyers mounted four distinct constitutional challenges in a single complaint. Future challengers have a ready playbook.
The Commerce Clause argument, if accepted, could be particularly sweeping. If courts agree that state AI laws effectively regulate AI products nationwide — because developers cannot build state-specific versions of large language models — then no state AI law can reach out-of-state developers at all.
The EU AI Act remains the only major jurisdiction with a comprehensive, legally operative AI framework. While the U.S. fights internally over whether states can regulate AI at all, the EU framework is already in implementation phase. Companies operating internationally will face divergent compliance requirements regardless of how U.S. litigation resolves.
What Organizations Should Do
1. Audit Your AI Systems Against the Narrowed Colorado Framework
Even with SB 24-205 suspended, SB 26-189 is advancing and will impose real obligations beginning January 1, 2027. Organizations that use automated decision-making technology affecting Colorado residents should inventory their systems now: identify which decisions are automated, document the methodology in plain language, and establish the 30-day adverse-decision notice workflow before the new law takes effect. Getting ahead of the transparency requirements is both good practice and good legal risk management.
2. Map Your Multi-State Regulatory Exposure and Prepare for Fragmentation
The Colorado situation is not unique — it is the leading edge of a fragmented national regulatory landscape. Legal and compliance teams should map which other state AI bills are advancing (notably in Illinois, Texas, and New York), identify commonalities in requirements, and build compliance infrastructure flexible enough to accommodate variation. Organizations that invest now in modular compliance architectures — where transparency, audit, and disclosure functions can be switched on per jurisdiction — will be far better positioned than those waiting for federal clarity that may not arrive.
3. Engage the Standard-Setting Process Before Litigation Forces Your Hand
xAI chose litigation; other companies have chosen engagement. The NIST AI Risk Management Framework, the EU-US Trade and Technology Council working groups, and industry bodies like the Partnership on AI are all active in defining what “reasonable care” and “impact assessment” look like in practice. Organizations that participate in these processes shape standards that courts and regulators subsequently adopt. If you wait until a law takes effect before defining your compliance methodology, you are defining it defensively under legal pressure — a far worse position than helping write the standard in advance.
The Bigger Picture: A Federal AI Vacuum
The deeper problem exposed by the Colorado episode is the absence of a federal AI framework. Congress has debated AI legislation for years without passing anything comprehensive. The executive branch has issued guidance and executive orders, but these lack the force of statute and can be reversed by the next administration.
In that vacuum, states have moved. Forty-five states introducing AI legislation in a single cycle is not a regulatory fad — it reflects genuine public demand for accountability in AI systems that affect housing, employment, healthcare, and financial services. The DOJ’s intervention does not fill the federal vacuum; it merely defends that vacuum by keeping states from filling it.
For companies, the uncertainty is itself a form of cost. Compliance programs built for one regulatory framework get dismantled when that framework collapses. Legal teams that tracked Colorado’s requirements for two years now have to pivot to SB 26-189’s narrower model, while also watching 44 other state legislatures.
The episode also raises a question that no court has yet definitively answered: at what point does a state AI law stop regulating AI deployed in the state and start regulating the AI industry nationally? The answer to that question will determine the outer boundary of state AI regulatory authority for years to come.
❓ Frequently Asked Questions
Q: Why did the DOJ side with xAI against Colorado’s AI law?
The DOJ did not simply support xAI’s position — it filed its own independent constitutional challenge focused on Equal Protection grounds. The DOJ argued that SB 24-205 both compels AI developers to engage in race- and sex-conscious engineering (to avoid disparate-impact liability) and explicitly authorizes differential treatment based on protected characteristics through its diversity carveout. Both positions aligned with the Trump administration’s broader opposition to diversity-focused regulatory requirements and with the December 2025 Executive Order creating the AI Litigation Task Force.
Q: Does the enforcement stay mean companies operating in Colorado don’t need to comply with SB 24-205?
Correct — for now. The April 27, 2026 court order suspended enforcement of SB 24-205 pending two conditions: the conclusion of the 2026 legislative session and a ruling on xAI’s preliminary injunction motion. Because the legislature passed SB 26-189 to replace the original law, the conditions for lifting the stay may never materialize. Companies should focus compliance planning on SB 26-189’s transparency requirements, which are set to begin January 1, 2027, rather than SB 24-205’s suspended framework.
Q: Could other states pass AI bias laws that survive the kind of constitutional challenge xAI brought?
Potentially, if they avoid the specific vulnerabilities xAI identified. A narrowly scoped law focused on transparency and individual rights — rather than statistical disparity liability — is harder to attack on Commerce Clause and First Amendment grounds. Avoiding diversity carveouts eliminates the Equal Protection target. Limiting applicability to in-state deployers rather than out-of-state developers reduces dormant Commerce Clause exposure. SB 26-189 itself represents an attempt to build exactly this kind of narrower, more defensible framework.
Sources & Further Reading
- Further Reading
- <ul>
- <li><a href=”https://www.justice.gov/opa/pr/justice-department-intervenes-xai-lawsuit-challenging-colorados-algorithmic-discrimination” target=”_blank” rel=”noopener noreferrer”>Justice Department Intervenes in xAI Lawsuit Challenging Colorado’s Algorithmic Discrimination Law — U.S. Department of Justice</a></li>
- <li><a href=”https://www.nortonrosefulbright.com/en-us/knowledge/publications/de3ad9de/xai-sues-doj-intervenes-enforcement-of-colorado-ai-act-suspended” target=”_blank” rel=”noopener noreferrer”>xAI Sues, DOJ Intervenes, Enforcement of Colorado AI Act Suspended — Norton Rose Fulbright</a></li>
- <li><a href=”https://btlaw.com/en/insights/alerts/2026/doj-intervenes-in-lawsuit-challenging-colorados-algorithmic-discrimination-law” target=”_blank” rel=”noopener noreferrer”>DOJ Intervenes in Lawsuit Challenging Colorado’s Algorithmic Discrimination Law — Barnes & Thornburg</a></li>
- <li><a href=”https://www.lawandtheworkplace.com/2026/05/major-developments-put-colorados-ai-law-on-ice-ahead-of-implementation/” target=”_blank” rel=”noopener noreferrer”>Major Developments Put Colorado’s AI Law on Ice Ahead of Implementation — Law and the Workplace</a></li>
- </ul>
🔗 Related Intelligence
Colorado’s AI Act: America’s Most Comprehensive State AI Law Faces Its Reckoning
US State AI Laws vs Federal Preemption: Colorado’s Delay, California SB 53, and the 2026 Clash
US State AI Laws in 2026: Colorado, California, Texas, and the Federal Collision
US DOJ AI Litigation Task Force: What the State-Law Preemption Battle Means Globally
