Tech / AI
After the DOJ Filed, Colorado Repealed Its AI Bias Law
Colorado's SB 24-205 required pre-deployment impact assessments, NIST-aligned risk controls, and attorney general discrimination reporting. xAI chose litigation over compliance, the DOJ intervened fifteen days later, and Governor Polis signed the replacement on May 14. No impact assessment for Grok is on the Colorado record.
Governor Jared Polis signed Senate Bill 26-189 on May 14, erasing Colorado's algorithmic discrimination guardrails before the original law reached its June 30 enforcement date.
SB 24-205, signed in 2024, required impact assessments within 90 days, NIST-aligned controls, and discrimination reporting to the attorney general. High-risk status attached to any system making consequential decisions in named domains: employment, lending, housing, healthcare. Colorado businesses were deploying Grok for several of them, pulling xAI into developer-tier compliance.
xAI filed case 1:26-cv-01515 in U.S. District Court in Denver on April 9, before the June 30 enforcement date arrived. The company's theory: Grok's outputs are protected expression, comparable to a video game's editorial choices or a search engine's judgment about relevance. An impact-assessment mandate, xAI argued, would force Grok to "abandon its disinterested pursuit of truth" and instead optimize for Colorado's preferred positions on racial equity.
The DOJ moved to intervene on April 24, fifteen days after the filing, the first such federal action against a state AI law. The brief cited Executive Order 14365, signed December 11, 2025, which established the AI Litigation Task Force and named Colorado's law by description. Colorado's attorney general stipulated to a stay on April 27; seventeen days later, Polis signed the replacement.
No impact assessment, discrimination report, or NIST-aligned audit for Grok is on the Colorado record. SB 26-189 removed the duty to prevent algorithmic discrimination, the 90-day assessment requirement, and the NIST-aligned mandate. Colorado's original statute was the only enacted domestic equivalent to the EU AI Act's high-risk tier; the replacement's notice obligations don't take effect until January 1, 2027.
The filing timeline reveals a replicable playbook: a preemptive filing brings DOJ in as co-counsel, and the legislature has nowhere to go. xAI's First Amendment theory treats fairness audits as compelled speech; the DOJ's equal-protection theory treats disparate-impact liability as compelled discrimination. Any lab following the same path now has a federal task force already pointed at state AI laws.
Case 1:26-cv-01515 remains active under a stay that covers "any legislation that may replace or amend" SB 24-205, extending xAI's injunction window to the replacement. SB 26-189 still requires developers to furnish deployers with documented statements of intended uses and known harmful uses, a compelled-disclosure target on the same First Amendment theory. If AG rulemaking concludes this fall, xAI has 28 days to file before the January 1, 2027 activation date.