Vermont AI Legislation in 2026

The Vermont Picture

Vermont occupies a distinctive position in the American AI policy landscape, and I would characterize its approach as governance-first rather than market-regulation-first. Years before most states acted, Vermont passed Act 132 of 2022, which created a Division of Artificial Intelligence inside the Agency of Digital Services and required the state to inventory every automated decision system it uses, propose a code of ethics, and report annually to the Legislature. That decision matters because it made Vermont one of the earliest states to treat its own use of AI as the first thing to discipline. The state effectively held itself to account before attempting to hold private builders to account, and that sequencing still defines the Vermont posture today.

The contrast with Vermont's private-sector record is stark, and builders should understand it clearly. The comprehensive ambitions, first H.711 in 2024 and then the H.340 and H.341 package in the 2025-2026 session, were among the most expansive attempts in the country to impose Colorado-style duties on developers and deployers of high-risk and dangerous AI systems. Every one of them stalled. Just as telling, the Vermont Data Privacy Act passed the Legislature in 2024 and was vetoed by Governor Scott, who warned it would make Vermont a national outlier and a hostile environment for small and mid-sized businesses. The Senate sustained that veto by a single vote. The lesson is that a passed bill is not a law, and the executive branch has been a consistent brake on broad mandates.

Where Vermont has actually moved, it has moved through narrow, high-salience harms. In 2024 it explicitly criminalized nonconsensual deepfake pornography by folding it into the existing nonconsensual imagery statute. In 2026 it enacted Act 75, requiring disclosure of AI-generated synthetic media in campaign communications inside the 90-day election window, and Act 101, which regulates mental health chatbots, requires notice when generative AI is used in patient communications, limits AI in utilization review, and establishes neural and mental data protections. These are not horizontal AI rules. They are targeted interventions in elections, intimate-image abuse, and health care.

For builders, the practical guidance follows directly from this pattern. If you sell AI into Vermont state government, expect to appear in the automated decision system inventory and to be measured against the state code of ethics, so document your system, its data, and its decision logic from the outset. If you operate political or election-adjacent generative tools, build conspicuous synthetic-media disclosure into the product now, because Act 75 is in force. If you deploy any chatbot or generative system in health, mental health, or human services, treat Act 101 as binding from July 1, 2026. And if your product touches intimate imagery, the criminal deepfake statute already applies. For now there is no comprehensive Vermont AI law, but I would not treat that as permanent.