United States Intervenes in xAI v. Colorado

xAI v. Weiser is a challenge to Colorado's law requiring that AI systems avoid "algorithmic discrimination." xAI challenged the law on First Amendment, Dormant Commerce Clause, and Fourteenth Amendment (vagueness) grounds.

xAI's complaint noted how troubled the passage of this legislation was. The governor, when signing the bill in 2024, expressed reservations about how its intended effect was implemented:

Laws that seek to prevent discrimination generally focus on prohibiting intentional discriminatory conduct. Notably, this bill deviates from that practice by regulating the results of AI system use, regardless of intent, and I encourage the legislature to reexamine this concept as the law is finalized before it takes effect in 2026.

xAI noted in its complaint some other problematic requirements:

7. SB24-205’s problems are many, but its most obvious constitutional defect appears in its first section. The law defines “algorithmic discrimination” as “any condition in which the use of an [AI] system results in an unlawful differential treatment or impact that disfavors an individual or group.” § 6-1-1701(1)(a). But the law then simultaneously promotes a form of “differential treatment” the State favors—discrimination intended to “increase diversity or redress historical discrimination.” § 6-1-1701(1)(b)(I)(B).

The differential treatment xAI refers to is that the law states that "algorithmic discrimination" does not include "The offer, license, or use of a high-risk artificial intelligence system by a developer or deployer for the sole purpose of: … (b) expanding an applicant, customer, or participant pool to increase diversity or redress historical discrimination …"

Perhaps not surprisingly, given both its antipathy to state regulation of AI and keen interest in anti-discrimination legislation, the federal government was paying attention and moved to intervene as co-plaintiff. The motion was granted the same day (Docket No. 16):

The Civil Rights Act of 1964 gives the United States an unconditional right to intervene in Equal Protection cases if the Attorney General certifies that it is a case of general public importance. See 42 U.S.C. § 2000h-2. X.AI LLC brings an Equal Protection claim and the Acting Attorney General has certified that this case is of general public importance. Doc. 1 187-201; Doc. 12-1. … The motion is GRANTED.

The U.S. government's complaint states two counts: "Violation of the Equal Protection Clause of the Fourteenth Amendment: Compelled Discrimination" and "Violation of the Equal Protection Clause of the Fourteenth Amendment: Authorized Discrimination." On the first count:

47. When a state law “compel[s] persons to discriminate against other persons because of race,” it is “a palpable violation of the Fourteenth Amendment.” Peterson v. City of Greenville, 373 U.S. 244, 248 (1963). Indeed, “if the . . . Government is prohibited from discriminating on the basis of race, . . . then surely it is also prohibited from enacting laws mandating that third parties . . . discriminate on the basis of race.” See Ricci v. DeStefano, 557 U.S. 557, 594 (2009) (Scalia, J., concurring) (citing Buchanan v. Warley, 245 U.S. 60, 78-82 (1917)). This logic extends to discrimination on the basis of sex and religion, which the Equal Protection Clause likewise forbids.
48. SB24-205 violates the Equal Protection Clause because it compels AI developers and deployers to discriminate based on race, sex, religion, or other protected characteristics. By requiring developers and deployers to prevent the “risk” of disparate outcomes based on demographic characteristics, SB24-205 effectively requires developers and deployers to expressly use demographic characteristics, including race, sex, and religion when building and using algorithmic models.

The second count refers to the exclusion for AI systems that "increase diversity" or "redress historical discrimination." The United States argues that

A state may draw so-called remedial racial classifications only if it can provide “evidence of past discrimination” by the state or private actors in the relevant area against the groups favored by the remedial classifications. City of Richmond v. J.A. Croson Co., 488 U.S. 469, 506 (1989) (emphasis omitted). Such evidence generally must show “past intentional discrimination” by the state.

However,

There is no evidence of past intentional discrimination by Colorado or by private algorithmic developers or deployers in the nascent industry of AI. Thus, Colorado’s efforts to prevent discrimination against "groups that, as a practical matter, may never have suffered from discrimination" in AI, "suggests that perhaps [its] purpose was not in fact to remedy past discrimination."

The Colorado law regulates the way AI systems are trained in general, which may be somewhat atypical. Texas law also prohibits someone from developing AI "in a manner that intentionally aims to incite or encourage a person to: (1) commit physical self-harm, including suicide; (2) harm another person; or (3) engage in criminal activity," but its law includes an intent requirement, an approach the Colorado governor believes is missing in the Colorado law. Other states regulate the user of the AI system, or give rights to the affected, not try to regulate how system itself works.¹

This case could illuminate whether and how it might be possible to regulate the development of AI systems – although it probably won't. All parties have asked for a suspension based on a stipulation that the Attorney General will not be promulgating rules or enforcing the law until the current legislative session, in which there is pending legislation to amend the law, has concluded. Reports are that any proposed amendment may eliminate the stand-alone affirmative duty to avoid algorithmic discrimination.