If you're tired of censorship and dystopian threats against civil liberties, subscribe to Reclaim The Net. A federal judge has frozen enforcement of Colorado’s first-in-the-nation AI law, the statute that would have required developers to police their own models for “algorithmic discrimination” and to inform the state of “foreseeable risks” before the rules took effect on June 30. Judge Cyrus Y. Chung signed off on a joint request from xAI and Colorado Attorney General Phil Weiser on April 27, putting the law on ice while state lawmakers draft a replacement. We obtained a copy of the order for you here. The order was filed in xAI v. Weiser. The state agreed not to enforce SB 24-205 against xAI, or to issue rules under it, until at least 14 days after the court rules on a forthcoming preliminary injunction motion. The June 16 scheduling conference was cancelled. The deadlines in the case are suspended. This is a significant retreat as Colorado spent two years insisting the law was a model for the country. It was the only state AI statute named in President Trump’s AI executive order last year. Now the state is asking a court to stop the clock while its own governor’s policy group drafts a bill to repeal and replace it. The law itself is the reason the climbdown looks the way it does. SB 24-205 told developers of “high-risk” AI systems they had to take “reasonable care” to prevent algorithmic discrimination, with one carveout that has done more work in the lawsuit than any other clause: the law exempts discrimination intended to “increase diversity or redress historical discrimination.” The state forbids one kind of discrimination by an algorithm. It permits, and arguably requires, another. The developer is left to figure out which is which, with the attorney general’s office deciding after the fact. xAI sued on April 9, calling the statute a First Amendment problem dressed up as consumer protection. The company’s complaint is more blunt than most filings of this kind. “SB24-205 is decidedly not an anti-discrimination law,” the company’s attorneys wrote. “It is instead an effort to embed the State’s preferred views into the very fabric of AI systems.” The argument is that Colorado isn’t regulating outputs neutrally. It’s choosing which viewpoints an AI model is allowed to produce, then enforcing the choice through “onerous policy, assessment, and disclosure requirements,” in the words of the Justice Department’s filing. The DOJ moved to intervene on xAI’s side, the first time the federal government has joined a constitutional challenge to a state AI regulation. Assistant Attorney General Harmeet K. Dhillon, who runs the Civil Rights Division, weighed in: “Laws that require AI companies to infect their products with woke DEI ideology are illegal.” You can take or leave the political register. The legal point underneath it is one anyone who cares about speech should take seriously. A state cannot tell a publisher, a newspaper, a search engine, or a chatbot which viewpoints its outputs must reflect. The First Amendment treats compelled speech as a near-cousin of censorship, and for the same reason: the government doesn’t get to write the script. Colorado’s law was vague enough to make almost any output a potential violation. The statute didn’t precisely define “algorithmic discrimination,” “foreseeable risks,” or what “reasonable care” looks like for a model with hundreds of millions of possible prompts. xAI’s complaint argues the statute is “unconstitutionally vague” and “invites arbitrary enforcement” because key terms are not defined. When a law is that loose, the chilling effect arrives before any enforcement does. Developers self-censor their models to stay on the safe side of a line the regulator hasn’t drawn yet. That self-censorship is the point, whether or not the law’s drafters intended it. A model that has to worry about Colorado’s interpretation of “disparate impact” will avoid topics, hedge answers, and decline questions. Colorado isn’t conceding any of this. The state lawmakers who backed the bill have pushed back. Rep. Brianna Titone, D-Arvada, a lead sponsor, told the Colorado Sun that SB 24-205 “is, and has always been, promoted as a policy to prevent and curtail discrimination for consequential decisions.” Rep. Manny Rutinel, D-Commerce City, accused the federal government of carrying water for Musk: “Coloradans deserve technology that works for everyone, not just billionaires.” Both responses sidestep the central issue. The question isn’t whether Colorado meant well. The question is whether a state can compel the speech of a software developer the way it just tried to. That power, once established, won’t stay limited to AI companies that the legislature dislikes. It will reach the next platform, the next publisher, the next set of opinions a future statehouse decides need correcting. The pause buys everyone time. Gov. Jared Polis’s AI policy group released a draft replacement bill on March 17, and the legislature is now preparing what would be the third round of amendments to a law that still hasn’t taken effect. Polis signed the original reluctantly in 2024, citing worries about the state’s tech sector. The original February effective date had already been pushed back to June 30 under industry pressure. This is a law that has been struggling to survive contact with reality from the day it was signed. If you're tired of censorship and dystopian threats against civil liberties, subscribe to Reclaim The Net. The post Judge Halts Colorado AI Law After First Amendment Challenge appeared first on Reclaim The Net.