AI Momentum 98.5 +1.5%GitHub Pulse 5503 +6.5%Capital Flow $204.0B +19039.3%Hiring Heatmap 0.0 -100.0%
Advertisement
Regulation & Policy

Colorado adopts rules for AI in hiring, lending and chatbots, setting up a court fight with xAI

Colorado has finished writing the rules for its two new AI laws. One governs software that helps decide who gets a job, a loan, an apartment or insurance. The other governs consumer chatbots. Both laws take effect January 1, 2027.

The state Department of Law, run by Attorney General Phil Weiser, adopted the rules on November 19. That step matters beyond Colorado. It is meant to start a court deadline for Elon Musk's xAI, which sued to stop the state's AI law and agreed to pause the fight until the rules were done.

For companies selling AI tools to lenders, employers, landlords and insurers, the result is an awkward planning problem. The compliance date is less than six weeks away. Whether the decision-making rules can be enforced on schedule may now depend on a federal judge.

What companies will have to do

The decision-making law covers software that materially influences a consequential decision, meaning its output is more than a trivial factor. Covered areas include education, employment, housing, lending, insurance, health care and essential government services.

Companies that use these tools must tell people beforehand that automated software is involved. After a denial or other bad outcome, they must explain in plain language, within 30 days, what role the software played. People can ask for their data, request corrections and seek a meaningful human review of the decision.

Companies that build the tools must hand their customers documentation on intended uses, the kinds of data used to train them, known limits and how humans should review results. Both sides must keep compliance records for at least three years.

According to trade publication PPC Land, which reviewed the 76-page adopted text, people who ask about a bad outcome can also learn the names of data brokers that supplied information, where the company knows them. The same report says human reviews must be acknowledged within 10 days and generally finished within 45.

This is a narrower law than the one Colorado passed in 2024. The replacement, signed in May, dropped the duty to take reasonable care against algorithmic discrimination. It also dropped required impact assessments and risk management programs, law firm Jones Walker noted. Only the attorney general can enforce it, and until 2030 companies generally get a chance to fix problems before facing enforcement.

The rules also eased some burdens on lenders during drafting. An October draft removed a requirement to pause a decision while a customer's correction or review request was pending, lawyers at Troutman Pepper Locke wrote. They still called the human review provisions very problematic for financial firms.

The chatbot rules

The second law targets consumer chatbots, with extra protections for teenagers. Operators must estimate users' ages and cannot rely on government ID as their only method, PPC Land reported. They must tell minors they are talking to an AI at the start of a session and at least every three hours.

Regulators will also look at whether operators stop chatbot characters from using titles like therapist or doctor. Features that reward minors for staying engaged, such as streaks, may be off limits. Larger operators, sorted into five tiers by monthly users, must file annual safety reports starting July 1, 2027, according to the same report.

For brands, the line matters. Bots limited to shopping, orders and customer support may fall outside the law. One that remembers past chats or offers constant emotional validation could drift into regulated territory, PPC Land's analysis suggested.

The court clock starts, and what it means

In April, a federal magistrate judge in Colorado paused the case at the joint request of xAI and the state. Under that deal, xAI agreed to file for a preliminary injunction, a court order blocking the law while the case proceeds, within 28 days of the state adopting rules, HR Dive reported.

In return, the attorney general agreed not to take enforcement action against xAI, including opening an investigation, until 14 days after the judge rules on that request, according to the joint court filing. That covers the 2024 law and its replacement. xAI has called the 2024 law "an effort to embed the State's preferred views into the very fabric of AI systems."

There is a catch in the timing. The court order does not define final adoption. If the November 19 vote counts, xAI's filing would be due around December 17, PPC Land calculated. If it means the date the rules take legal effect, the deadline slides later.

The rules are not yet in force. The department asked the attorney general for a formal legal opinion on November 20. The rules and that opinion then go to the Secretary of State, where the filing was listed as pending, and they become final 20 days after publication.

One more wrinkle: the pause does not name the chatbot law, according to PPC Land. So the teen-safety rules may face a different path than the rules on hiring and lending software.

Colorado's 2024 law was one of the first in the country aimed at bias in AI-driven decisions. Its replacement now has detailed operating rules. Vendors that sell nationally may find it simpler to build one set of disclosure and appeal tools than a Colorado-only version.

Buyers should not read the court fight as a reason to wait. Discrimination claims under Colorado's existing anti-discrimination law do not depend on the new rules taking effect, Jones Walker pointed out. Under the new law, contract clauses that shield a company from liability for its own discriminatory use of the software are void.

The next date to watch is xAI's filing. A ruling for xAI would freeze one of the most closely watched state AI laws in the country. A ruling for Colorado would give other states a tested template to copy.

Written by AI from the public sources linked below.