An act relating to regulating developers and deployers of certain automated decision systems
H.340 would create a new chapter in Vermont’s consumer protection law regulating the development and use of automated decision systems when they are used in “consequential decisions.” The bill defines consequential decisions broadly to include decisions affecting education, employment, housing, utilities, family planning, health care, financial services, law enforcement, public benefits, and civil-rights accommodations. It makes it unlawful for a developer or deployer to use, sell, or share an automated decision system that produces algorithmic discrimination, and it sets out detailed obligations for notice, explanation, consumer appeal, human review, whistleblower protections, independent audits, reporting, and risk management.
The bill requires deployers to notify consumers when automated systems are used, explain the system’s role and the reasons for decisions, and provide a process for appeal with meaningful human review. Developers and deployers must file reports with the Attorney General before deployment and annually or after substantial changes, and they must submit independent audits showing the system does not produce algorithmic discrimination. The Attorney General would also maintain a public database of redacted reports and audits, adopt rules for implementation, and enforce violations as unfair and deceptive acts under the Vermont Consumer Protection Act. The bill would take effect July 1, 2025.
H.340 would add a new artificial intelligence regulatory framework to Title 9 of Vermont law and tie violations to the Vermont Consumer Protection Act. It would impose new compliance duties on businesses that develop or use automated decision systems in high-stakes settings, including documentation, consumer disclosures, appeal procedures, independent audits, and risk-management programs aligned with NIST AI risk guidance. The Attorney General would gain new rulemaking, investigative, and enforcement authority, and consumers harmed by violations would have access to consumer-protection remedies.
Based on the bill text and the absence of recorded committee testimony or votes in the provided materials, the overall posture appears precautionary and consumer-protective rather than controversial in the record available here. The bill is framed as a response to algorithmic discrimination and emphasizes transparency, accountability, and human oversight. Because no committee transcripts or vote history were provided, there is no documented split in sentiment to report from the available context.
The main points of potential contention are the bill’s breadth and compliance burden. Businesses and technology developers may object to the extensive reporting, audit, and disclosure requirements, the limits on using systems before independent audits are completed, and the possibility of joint and several liability if responsibility is not allocated by contract. Another likely area of debate is the scope of covered decisions and protected traits, as well as the requirement for meaningful human review and public reporting to the Attorney General. Supporters would likely emphasize consumer protection, anti-discrimination safeguards, and transparency, while critics may focus on cost, administrative complexity, trade secret concerns, and the feasibility of auditing complex AI systems.