New Hampshire AI Compliance Guide

Building a compliance program under No AI-specific law requires a sequential, documented approach — not a single-event audit. While New Hampshire does not yet have a dedicated AI law in effect, the compliance program described here applies the federal framework that currently governs AI use in this state and positions your business for state law adoption when it occurs. The steps below are sequenced in order of legal priority, not organizational convenience.

Step one is an AI inventory — a documented record of every AI system your organization uses, including AI embedded in third-party software such as CRM assistants, HR platforms, underwriting engines, customer service bots, and content tools. For each system, the inventory should capture: the vendor and model version; the specific decisions or recommendations the system outputs; whether those outputs influence consequential decisions affecting individuals in areas like employment, credit, insurance, housing, or healthcare access; and who within your organization is responsible for overseeing the system. New Hampshire businesses often discover during this exercise that they have more AI touchpoints than compliance leadership realized — particularly when embedded AI in enterprise software is counted separately from deliberate AI deployments. The inventory is the foundation on which every subsequent compliance step is built.

Step two is risk classification. Once inventoried, each AI system must be evaluated against No AI-specific law's scope criteria to determine whether it triggers compliance obligations. No state AI law. Legislature monitoring federal developments. High-impact AI systems — those that influence access to employment, credit, housing, insurance, healthcare, or government services — generate the most extensive obligations: written impact assessments, bias and fairness testing across protected demographic groups, disclosure notices to affected individuals, human-review pathways for adverse decisions, and records retention sufficient to reconstruct each automated decision. AI systems with limited individual impact generate narrower obligations focused on disclosure and documentation. Classifying each system accurately is essential because misclassification — treating a high-impact system as low-impact — creates exactly the enforcement exposure the compliance program is designed to avoid.

Step three is disclosure implementation. The disclosure requirement under No AI-specific law takes effect N/A. For each high-impact AI system, disclosure means notifying the affected individual — in plain language, before the AI decision becomes final — that an automated system materially influenced the outcome. Disclosure notices must be accessible, not buried in terms of service, and specific enough to be meaningful. They must be paired with a mechanism for the individual to request human review or contest the decision. Businesses should also add a public-facing AI usage statement to their website, update their privacy policy to reference AI systems and the data they process, and ensure consumer-facing disclosure language is reviewed by counsel for compliance with New Hampshire's specific statutory requirements.

Step four is technical controls. The most important technical controls for New Hampshire AI compliance are: first, audit logging — per-decision records that capture the inputs, model version, output, and the identity of any human reviewer, retained for at least three years or the applicable statute of limitations; second, human-review checkpoints — a defined process by which an individual can escalate an AI-driven adverse decision to a human decision-maker with authority to override; third, data minimization — limiting the personal data sent to AI systems to what is operationally necessary, reducing both AI risk and data-protection exposure; and fourth, content provenance — for businesses generating AI-created content, metadata or labels that satisfy New Hampshire's disclosure requirements around AI-generated text, images, audio, and video. Documented technical controls are a recognized mitigating factor in enforcement proceedings.

Step five is vendor management. Every third-party AI tool your organization uses creates compliance obligations that flow back to you as the deployer. Before deploying or renewing a vendor AI tool, conduct documented due diligence that covers: whether the vendor has performed bias and fairness testing on the model and can share results; whether the vendor's contract includes a data-processing agreement covering AI-specific obligations such as training-data use, sub-processor disclosure, and retention limits; and whether the vendor provides indemnification for AI-law-specific violations. Update existing vendor agreements for any high-impact AI tools already in production. Vendors that cannot provide basic documentation of their AI system's testing and compliance posture should be treated as high-risk — deploying their tools without that documentation creates documented exposure for your organization that cannot be shifted to the vendor after the fact.

Step six is ongoing monitoring, staff training, and program maintenance. AI compliance is not a one-time audit; it is a continuous operational function. Each high-impact AI system should be re-evaluated for bias and risk at least annually and after any material model update or training-data change. Compliance logs should be reviewed monthly to verify that disclosure and human-review pathways are functioning correctly. All employees who interact with AI systems in consequential workflows must be trained on New Hampshire's disclosure obligations and how to escalate compliance concerns. Designate an AI compliance owner — a named individual responsible for maintaining the inventory, tracking regulatory updates, and owning the organization's relationship with the relevant New Hampshire enforcement authority. Standing up this function before N/A ensures your organization is compliant from day one of enforcement.

Explore More for New Hampshire

Sources verified against official .gov filings · Last verified Apr 22, 2026.

• ↗gencourt.state.nh.ushttps://www.gencourt.state.nh.us/
• ↗ncsl.orghttps://www.ncsl.org/research/telecommunications-and-information-technology/s…