Rhode Island AI Legislation in 2026

The Rhode Island Picture

Rhode Island has chosen a deliberately narrow, harm-specific path on AI rather than the sweeping, cross-sector framework that Colorado and Texas have adopted. The two pieces of binding law the General Assembly has actually put on the books target concrete, well-understood harms: AI-generated deepfakes in political campaigns and nonconsensual synthetic intimate imagery. The election law, S 0816, signed by Governor McKee on July 2, 2025, is a disclosure regime, not a prohibition. It bars undisclosed synthetic media in candidate, party, and PAC communications within 90 days of an election, but it expressly preserves satire, parody, and any AI content that carries a clear label. Rhode Island regulates deception, not the technology itself.

The comprehensive ambition exists, but it keeps failing. Rhode Island has now introduced and lost at least three serious attempts at a Colorado-style high-risk AI law: H 7521 and S 2888 in 2024, and S 0627 in 2025, which was held for further study and died in committee. Each version would have imposed familiar obligations on developers and deployers of high-risk systems used for consequential decisions: reasonable care against algorithmic discrimination, documented risk management, impact assessments, consumer notice, a right to appeal, and Attorney General oversight. None survived. I read this as a legislature interested in the framework but not yet willing to enact one, in part because of First Amendment and overbreadth concerns. Anyone planning around Rhode Island law should treat a general high-risk AI statute as plausible but not imminent.

The most active live front for builders is health care and AI companions. H 5172 would require human physician review of any AI-influenced adverse coverage determination and five-year retention of AI decision records, and S 2195, introduced in the 2026 session with a proposed January 1, 2027 effective date, would impose safety and emotional-disclosure duties on AI companion chatbots. Neither is law yet, but both signal where the next enforceable obligations are likely to land. If you deploy AI in Rhode Island health insurance decisioning or operate consumer-facing companion products, design now for human-in-the-loop review, durable audit logging, self-harm escalation handling, and explicit this is not a human disclosures.

For practical compliance today, the rules are clear and limited. If you produce political content in Rhode Island, label AI-generated or AI-altered media or keep it outside the 90-day pre-election window, and understand that targeted candidates can sue. If your product can generate intimate imagery of real people, the synthetic-image criminal exposure under the updated statute is real. Beyond those two areas, Rhode Island currently relies on its executive-branch AI Task Force under Executive Order 24-06 rather than statute to shape state AI policy, so the prudent move is to build to the stricter emerging multistate norm, because that is the direction every failed Rhode Island bill has pointed.