Colorado’s AI Act: Still Standing

9 Min Read By: Michael Simon

In Brief

  • The continuing technology industry efforts to defeat the Colorado AI Act came to a head in an August 21–27 special legislative session called by the governor.
  • Much drama ensued in Denver as the proponents of four competing bills and over 150 lobbyists clashed over what many see as the future of AI regulation not just in Colorado, but in the United States as a whole.
  • Attempts to add further provisions to the AI Act, such as joint and several liability, failed, but so did attempts to completely gut it.
  • Ultimately, the AI Act’s proponents agreed to delay implementation until June 30, 2026, which was just slightly past their original proposed compromise three-month delay.
  • This isn’t over: Colorado legislative leaders, both proponents and opponents of the AI Act promise to resume attempts to reach a compromise draft next January, when the legislative cycle begins again.

Looking back on when I first wrote about the regulatory circus surrounding Colorado’s groundbreaking AI Act in August, I predicted the upcoming special session would add yet another chapter to this ongoing regulatory saga. We had no idea just how dramatic that chapter would be.

The August 21–27 special session didn’t just add a chapter—it delivered a full-blown political thriller complete with backroom deals, lobbying blitzes, late-night negotiations, shouting in the Capitol halls, and a climactic Monday morning collapse that left even seasoned observers stunned. The headlines rushed to try to turn this into a victory by the tech lobby:

Yet, this was no victory for Big Tech, because even though no one thought it could win, the Colorado AI Act is still standing.

The Setup: Four Bills, 150+ Lobbyists, and a Ticking Clock

Colorado Governor Jared Polis included the AI Act in his special session call, ostensibly to address budget shortfalls but also to give lawmakers one more chance to modify or delay the law before its February 1, 2026, effective date.

Four different bills emerged targeting the AI Act, each taking radically different approaches:

  • SB 25B-004: The “Colorado Artificial Intelligence Sunshine Act”—a complete rewrite focusing on transparency and disclosures, with Colorado Senate Majority Leader Robert Rodriguez one of the lead sponsors
  • SB 25B-008: Total repeal, replaced with “technology-neutral anti-discrimination” language
  • HB 25B-1009: Dramatic scope reduction (employment and public safety only) plus delays and exemptions
  • HB 25-B1008: Another near-total repeal with minimal disclosure requirements

But the real story was about the lobbying tsunami. An Axios Denver analysis of state records found that “more than 100 companies and organizations hired roughly 150 lobbyists to shape Rodriguez’s bill,” with Amazon, major health care companies, and a coalition of tech CEOs among those that hired multiple lobbyists.

150 lobbyists. For a six-day special session. In a state with just under 6 million people.

Act I: The Ambitious Rewrite

Rodriguez started with genuine ambition. His initial SB 4 was a comprehensive thirteen-page replacement that would have created the “Colorado Artificial Intelligence Sunshine Act.” The bill maintained consumer protections while streamlining requirements—exactly the kind of thoughtful compromise that seemed achievable.

The initial engrossed version included new definitions of “algorithmic decision systems,” developer disclosure requirements, individual data rights, and crucially, joint and several liability provisions for developers and deployers. It represented months of behind-the-scenes negotiations aimed at threading the needle between consumer protection and industry workability. It also included delays in the effective dates of key consumer rights provisions, back to May 1, 2026.

Act II: The Sunday Night Deal That Wasn’t

By Sunday night of the special session, August 24, it appeared lawmakers might have reached a deal. “Top Democrats told their colleagues that they had crafted the framework of an agreement,” the Colorado Sun reported. Rodriguez had made several key concessions, including removing from SB 4 a controversial requirement that deployers provide to individuals given an adverse decision from AI a list of the twenty personal characteristics of those people that most influenced the decision. He also offered a three-month delay in implementation of the AI Act.

But then came the liability provision that broke everything.

SB 4 would have amended the AI Act to add a provision creating joint and several liability for developers and deployers for any AI system violating the law, but with some safe harbor protections for some circumstances. During debate on the Senate floor, Rodriguez removed the safe-harbor protections and thus created what the industry, and its horde of lobbyists, viewed as unacceptable legal exposure.

Act III: The Monday Morning Meltdown

By Monday morning, the compromise started to fall apart, mainly over the degree of liability AI developers and deployers should face when their technology leads to discrimination.

The Colorado State Capitol became a pressure cooker. As described by the Colorado Sun, negotiations over the law had “rocked the Capitol since lawmakers returned to the building Thursday for a special session, with shouting at times filling the halls outside of the House and Senate. Democrats have been disagreeing with each other. Lobbyists have been livid. Confusion, anger and rumors have spread like a virus.”

State Senator Julie Gonzales captured the moment perfectly: “All thirty-five of us in this building know that we too have witnessed the stunning brunt of AI leverage.”

Rodriguez tried to save his compromise by pulling the liability amendment, but it was too late. “Business, consumer protection advocates, labor and educators came together, but big tech didn’t like the bill because they don’t like the liability,” Rodriguez said.

State Representative Brianna Titone, a lead sponsor of both the original AI Act and SB 4—who withdrew her sponsorship of SB 4 on Monday night—summed up the frustration: “we had a good thing going” with the potential compromise. Titone aimed her frustration at the tech companies, accusing them of being unreasonable in how much influence they were seeking over the drafting process.

The Great Pivot: From Rewrite to Delay

By Monday, August 25, facing certain defeat, Rodriguez was forced to make a strategic retreat. He told reporters that it had become “impossible” to find a compromise that would have worked for everyone. In a remarkable legislative maneuver, Rodriguez gutted his comprehensive thirteen-page rewrite and replaced it with a simple find-and-replace operation: every instance of “February 1, 2026” became “June 30, 2026.” The entire scope of the special session’s AI work boiled down to a four-month delay.

After hours of work from the captive legislators who could not leave until there was a resolution on the AI Act, the last vestiges of hope for a deal were gone. Rodriguez opted to release the legislature from what was starting to feel like a hostage situation.

The Real Victory: What Didn’t Happen

Here’s what the breathless headlines about “Big Tech wins” and “AI law gutted” missed: nothing fundamental changed. The Colorado AI Act remains the nation’s first comprehensive AI regulation. All its core provisions survived intact:

  • Risk assessments and impact assessments: still required
  • Transparency and disclosure requirements: unchanged
  • Consumer appeal rights: preserved
  • Developer documentation obligations: intact
  • Attorney general enforcement authority: untouched
  • Anti-discrimination protections: fully maintained

The industry mobilized more than 150 lobbyists, four different repeal/replacement bills, intense pressure campaigns, and months of negotiations. The result? A four-month delay—even though the AI Act proponents were already proposing what was effectively a three-month delay.

That’s not a victory for opponents—that’s a stunning testament to the law’s resilience.

The Broader Context: An AI Regulatory Law That Refuses to Die

Step back and look at the full timeline since Governor Polis signed the AI Act in May 2024:

  • May 2024: Polis signs with reservations, calls for revisions
  • June 2024: Unprecedented letter from the governor, the attorney general, and the law’s own sponsors calling for changes
  • 2025 Regular Session: SB 318 amendment effort fails
  • May 2025: Last-ditch filibuster by Titone prevents delay
  • Summer 2025: Federal preemption efforts in Congress fail
  • August 2025: Massive special session lobbying blitz achieves only a delay

At every turn, when given the opportunity to gut, delay, or kill the law, Colorado’s political system has ultimately chosen to preserve it. Even Rodriguez’s final floor speech, while announcing the delay, was a passionate defense of the law’s principles: “Should a company whose AI system determines who gets hired and promoted, how much tenants pay for rent, and who receives medical care ever be held to account?”

The Industry’s Pyrrhic Victory

Make no mistake: the tech industry got what it asked for in the short term. The delay provides breathing room and another chance to lobby for changes during the 2026 regular session. Organizations such as the Colorado Technology Association celebrated, with CEO Brittany Morris Saunders stating, “By extending the timeline, we now have the opportunity to work collaboratively on practical solutions that strengthen consumer trust, safeguard jobs, and preserve Colorado’s competitiveness.”

But this “victory” came at enormous cost. The industry revealed the extent of its political muscle, burning goodwill and reinforcing narratives about Big Tech’s outsized influence. Titone captured this perfectly: “They didn’t want to be responsible for the products that they make. And that should be alarming to everybody.”

More importantly, the industry failed to achieve any of its core substantive goals. It obtained no scope reductions, no big carve-out exemptions, no liability safe harbors. No repeating its success in watering down a law until it is simply an empty, sharp suit that contains only provisions that are nearly impossible to prove, which is arguably what happened to the Texas Responsible AI Governance Act (“TRAIGA”). The AI Act the industry will face on June 30, 2026, is identical to what it would have faced on February 1.

What’s Next

The 2026 regular session will likely be the last, best chance for the tech industry to secure meaningful changes to Colorado’s AI Act. But the political dynamics have shifted decidedly against it:

  1. Proven Resilience: The law has survived every major challenge, demonstrating its political staying power.
  2. Implementation Reality: With just months before the new deadline, wholesale changes become practically impossible.
  3. National Attention: Other states are watching Colorado as a model, creating pressure to maintain the law’s integrity.

State Senator Jeff Bridges offered perhaps the most realistic assessment: “There are folks involved in this that have taken Colorado’s first-in-the-country law and worked really hard to find a path forward. . . . We can get this done. This delay buys us that time.”

Practical Advice

For businesses developing or deploying AI to Colorado consumers, don’t let political uncertainty delay practical preparation. The core compliance obligations haven’t changed one bit. Businesses still need:

  • AI impact assessments, which take significant time and resources to design properly
  • Risk management programs
  • Vendor agreement reviews to ensure AI service providers can support compliance obligations
  • Staff training on AI governance
  • Process documentation

The only thing that has changed is that businesses have four extra months to get ready.

The Bigger Picture: Colorado’s Regulatory Resilience

The August special session ultimately demonstrated something remarkable: when it comes to protecting consumers from AI discrimination, Colorado’s AI Act has achieved an almost gravitational pull. Multiple attempts to significantly alter or delay it have failed. Even when facing unprecedented industry pressure, the political system defaulted to preservation rather than destruction.

This isn’t just about AI regulation; it’s about how democratically enacted consumer protections can withstand concentrated corporate opposition. In an era when regulatory capture often dominates policy discussions, Colorado’s AI Act represents something different: a law that, once passed, has proven remarkably difficult to undo.

The regulatory saga continues, but the plot has become clearer. This isn’t a story about a law struggling to survive—it’s about a law that has found its footing and refuses to be moved.

By: Michael Simon

MORE FROM THIS AUTHOR

Connect with a global network of over 30,000 business law professionals

18264

Login or Registration Required

You need to be logged in to complete that action.

Register/Login