Elon Musk’s AI company filed a federal lawsuit on Thursday asking a judge to block Colorado from enforcing a law that would let the state dictate what Grok can and cannot say.
The complaint, lodged in the US District Court for the District of Colorado against Attorney General Philip Weiser, calls Senate Bill 24-205 unconstitutional on First Amendment, Dormant Commerce Clause, and Equal Protection grounds. xAI wants the whole thing thrown out before it takes effect on June 30.
We obtained a copy of the lawsuit for you here.
SB 24-205 defines “algorithmic discrimination” as any AI output that results in “unlawful differential treatment or impact that disfavors an individual or group” based on protected characteristics.
Reclaim Your Digital Freedom.
Get unfiltered coverage of surveillance, censorship, and the technology threatening your civil liberties.
But the law then carves out an exemption for discrimination designed to “increase diversity or redress historical discrimination.” The state, in other words, built a law that bans one kind of differential treatment while explicitly blessing another. The distinction rests entirely on whether Colorado approves of the reason.
That’s a content and viewpoint distinction written into statute. xAI’s complaint argues the law “compels Plaintiff xAI to alter Grok, forcing Grok’s output on certain State-selected subjects to conform to a controversial, highly politicized viewpoint.”
The company calls the measure “an effort to embed the State’s preferred views into the very fabric of AI systems” and says it would force developers “to distort their AI models to seek and output progressive ideology instead of the truth.”
The constitutional argument about the bill itself has real weight. A law that regulates AI outputs based on whether the resulting discrimination serves the state’s preferred goals is the kind of viewpoint-based speech regulation that courts subject to the highest level of scrutiny.
The law’s scope is enormous. A “high-risk” AI system is defined as one that “makes, or is a substantial factor in making, a consequential decision” in areas including employment, housing, education, healthcare, and financial services.
The definition of “substantial factor” is breathtakingly broad, covering “any use of an artificial intelligence system to generate any content, decision, prediction, or recommendation concerning a consumer that is used as a basis to make a consequential decision.”
If someone uses Grok to draft interview questions or summarize a stack of CVs, that’s enough. The AI doesn’t have to make the hiring call itself. It just has to touch the process.
And the law applies wherever a single Colorado resident might be affected. xAI is incorporated in Nevada, headquartered in California, and has no offices in Colorado. The complaint argues that SB 24-205 regulates development and deployment activities that happen entirely outside the state, in violation of the Dormant Commerce Clause.
Colorado’s own political leadership hasn’t been able to settle on whether this law is a good idea. Governor Jared Polis signed it in May 2024 “with reservations,” warning that “Government regulation that is applied at the state level in a patchwork across the country can have the effect to hamper innovation and deter competition in an open market.”
Polis flagged something specific for the speech analysis. He noted that “[l]aws that seek to prevent discrimination generally focus on prohibiting intentional discriminatory conduct,” but SB 24-205 “deviates from that practice by regulating the results of AI system use, regardless of intent.”
That’s a significant admission from the person who signed the bill into law. The measure creates liability not for intentional bias but for statistical outcomes, and only the outcomes Colorado doesn’t like.
Attorney General Weiser himself called the bill “really problematic” in August 2025 and said it “needs to be fixed.”
A joint letter in May 2025, signed by Polis, Weiser, two US representatives, a US senator, and Denver’s mayor, asked the state legislature to delay the law until January 2027 so they could fix the problems.
The legislature instead just pushed the effective date to June 30, 2026, and left everything else untouched.
A March 2026 working group proposal to strip the algorithmic discrimination requirement hasn’t been introduced as a bill by any legislator. So the law stands as written.
The First Amendment argument has multiple layers. xAI argues that every choice a developer makes when building an AI model is an expressive activity, from selecting training data to writing system prompts to calibrating guardrails.
The complaint cites the Supreme Court’s 2024 decision in Moody v. NetChoice, which held that social media platforms engage in speech when curating content, and quotes the Court’s observation that on “the spectrum of dangers to free expression, there are few greater than allowing the government to change the speech of private actors in order to achieve its own conception of speech nirvana.”
The complaint also contends SB 24-205 burdens users’ right to receive information.
The argument is straightforward enough. If the law forces developers to alter model outputs so they don’t produce disfavored statistical patterns, then users get sanitised answers instead of whatever the model would have generated without state interference.
You don’t have to agree with everything xAI does to recognize that a state compelling a specific ideological adjustment to AI training data is a genuinely alarming precedent for speech.
There’s a vagueness problem, too. SB 24-205 prohibits “algorithmic discrimination” and exempts discrimination that redresses “historical discrimination,” but never defines “historical discrimination.”
It leaves that to the Attorney General to figure out through rulemaking. The law creates a $20,000-per-violation penalty for noncompliance, gives the AG exclusive enforcement authority, and has no private right of action. That concentration of definitional and enforcement power in one office, combined with terms that nobody can pin down, is the kind of arrangement that chills speech even before a single enforcement action.
