Senate Judiciary advances bill to criminalize nonconsensual and AI-generated intimate images, expand CSAM coverage

Senate Judiciary on Monday advanced Senate Bill 2-88, a comprehensive update to Colorado law addressing nonconsensual intimate images and artificially generated child sexual abuse material (CSAM). Sponsors and a coalition of prosecutors, child-advocacy centers and law-enforcement officials argued the bill closes legal loopholes created by rapidly improving AI image-generation tools; defense and civil-liberties groups warned the measure as drafted raises constitutional and evidentiary risks.

Senator Emil Rodriguez, the bill sponsor, said the measure combines criminal and civil tools: it criminalizes distribution and possession of nonconsensual explicit images when certain elements are satisfied, creates a civil cause of action with injunctive relief and damages for victims, and updates CSAM definitions so computer-generated images that appear to depict children can be prosecuted. Rodriguez and supporters said the bill mirrors model language used in other states while adding provisions tailored for Colorado.

Proponents at the hearing included Jessica Dotter from the Colorado District Attorneys Council, Jeff Reester from the Department of Law, commanders from the Colorado Internet Crimes Against Children Task Force, and executives from children's advocacy centers. ‘‘These images are not harmless. They're violations,’’ said Will Braunstein, executive director of the Denver Children's Advocacy Center, describing instances where students'faces were used in explicit AI images and circulated in schools.

Opponents included the ACLU of Colorado, the Office of the State Public Defender, and the Colorado Criminal Defense Bar. Their witnesses urged tighter drafting, noting key concerns: (1) section 2 as introduced appears to reach "complete fakes" that do not rely on real victims, which raises free-speech and overbreadth questions under U.S. Supreme Court precedents (including the Ashcroft line of cases); (2) technical and evidentiary challenges in reliably distinguishing AI-generated images from photographs, which complicates foundation for admission and mens rea; and (3) the bill departs in places from the "model law" referenced by stakeholders (a Public Citizen draft circulated to DA offices), particularly on the CSAM/obscenity elements.

The committee adopted three sponsor amendments requested in stakeholding: L001 (broadcaster liability protections for third-party content taken down in good faith), L002 (clarifying the bill's declaration/intent language), and L003 (technical cleanup requested by prosecutors). After amendments, Majority Leader Rodriguez moved the bill to the Committee of the Whole; the motion passed on a 7-0 recorded vote. At the sponsor's request the bill was not placed on the consent calendar.

Key points and fault lines

- Civil remedies: Sponsors described a civil cause of action that would permit victims to seek injunctive relief and recover actual or liquidated damages (sponsor stated a minimum liquidated amount of $150,000 in testimony), and to prohibit future disclosures. Opponents asked for clarity on notice, burdens of proof and First Amendment limits on damages.

- "Complete fakes" vs. victim-based imagery: Defense witnesses said the model law focuses on deepfakes that superimpose a real person's likeness; section 2 as drafted appears to reach images entirely synthesized by AI that do not depict a real person, raising constitutional questions about criminalizing speech without a human victim.

- CSAM and obscenity: The bill expands the definition of exploitative child material to include AI-created images. Several defense witnesses argued that U.S. Supreme Court precedent (including Ashcroft v. Free Speech Coalition) limits criminalization of wholly virtual child pornography absent an obscenity analysis, making this drafting area legally fraught.

- Investigative/evidentiary complexity: Law-enforcement witnesses described the increasing difficulty of determining whether an image is AI-generated or tied to an identifiable victim. Investigators rely on metadata, device forensics and other indicators, but panelists said the pace of technological change complicates standard investigative workflows.

What happens next

The bill moves to the Committee of the Whole. Sponsors and stakeholders said they will continue negotiations on second-reading amendments to narrow or clarify contested language, particularly the provisions governing wholly AI-generated images and the civil damages mechanism.