New Hampshire Drew a Line Around AI and Licensed Professional Services — But Nobody Knows Where the Line Is Yet
Published November 13, 2025 · By The Crossing Report
New Hampshire Drew a Line Around AI and Licensed Professional Services — But Nobody Knows Where the Line Is Yet
The professional licensing question at the heart of AI in professional services has finally become law — in New Hampshire.
New Hampshire SB 640, which passed committee the week of March 9, 2026, states that AI systems may not "provide licensed professional services" without a licensed professional's meaningful oversight. It covers law, accounting, medicine, and related professions.
That's the good news for professional services firm owners: the standard is explicit. The bad news: the standard is also completely undefined. "Meaningful oversight" is not defined in the bill. What that means in practice will be determined by enforcement, litigation, or regulatory guidance that doesn't exist yet.
This is a law that sets the rule before the rule has been interpreted. Your job is to figure out how to comply with something that hasn't been tested.
Summary
New Hampshire SB 640 establishes that AI may not provide licensed professional services without meaningful oversight from a licensed professional — covering law, accounting, medicine, and related fields. For firm owners, this creates both a compliance standard and an open question: what counts as 'meaningful oversight' when AI drafts the document or writes the client communication? That definition will shape professional liability cases for the next decade.
What the Law Is Trying to Prevent
The target is clear even if the standard isn't.
New Hampshire SB 640 is designed to prevent AI from acting as a substitute for a licensed professional — answering legal questions, providing tax advice, delivering accounting opinions, or making clinical recommendations — without a licensed person actually in the loop. The concern: AI that appears to provide professional services while the licensed human is either absent from the specific interaction or performing a perfunctory, non-substantive review.
This is the difference between:
- AI assistance: A licensed attorney uses AI to draft a contract clause, reviews it, modifies it, and delivers it with professional judgment applied. The human is meaningfully in the loop.
- AI provision: An AI system generates a contract, answers a client's legal question, or delivers a compliance opinion directly to a client based on automated processing. The human reviews 200 similar outputs in 20 minutes without actually reading them. The license is nominally attached but no professional judgment was applied.
The bill doesn't say AI can't help. It says AI can't be the licensed professional.
The Undefined Standard Is the Real Compliance Problem
"Meaningful oversight" is the operative term. The bill doesn't define it. Here's why that matters:
For law firms: If an AI tool drafts a contract clause and a partner glances at it for 30 seconds before sending, is that meaningful oversight? If an AI answers a client's preliminary question via a website chatbot and a licensed attorney is notified of the interaction but doesn't personally review it, is that meaningful? Nobody knows yet.
For accounting firms: If AI prepares a draft tax return and a CPA spends 10 minutes reviewing a 40-page return, does that meet the standard? If an AI tool provides a client with preliminary tax guidance through an automated portal and a CPA is listed as responsible, is the oversight meaningful?
For consulting firms: This one is more ambiguous — consulting is licensed in some specializations (financial advisory, HR consulting in certain contexts) but not others. Firms in licensed consulting specializations need to watch this more carefully than general management consultants.
The enforcement history on unauthorized practice of professions is the best precedent available. Courts have generally asked: did the output require professional judgment? Did a licensed professional actually apply that judgment to this specific output before it reached the client? NH SB 640 is applying that same inquiry to AI.
The Conservative Compliance Posture
In the absence of defined standards, here is the defensible position for any professional services firm providing services to New Hampshire clients:
Document the review, not just its existence.
It's not enough to say a licensed professional reviewed AI output before delivery. The review should produce evidence that it happened: a timestamp, an annotation, a version with tracked changes, a logged approval. If the review is indistinguishable from rubber-stamping, it will eventually be characterized that way.
Make the review specific to the output.
A blanket policy of "a partner reviews all AI-generated content" is weaker than specific review of each client deliverable. The meaningful oversight question is about the specific interaction — did a licensed professional apply judgment to this output before this client received it?
Separate AI-assisted work from AI-delivered work.
The cleanest position is the distinction in the billing or engagement process: AI helps prepare, a licensed professional delivers. When AI and the professional are indistinguishable in client-facing delivery, the oversight question becomes harder to answer.
Apply existing unauthorized practice standards.
The professional licensing standards your state already applies to unauthorized practice of professions are the best available guide to what "meaningful oversight" will mean. If you wouldn't let a non-licensed person sign off on a deliverable without licensed review, the standard for AI should be the same.
Why This Law Matters Beyond New Hampshire
New Hampshire SB 640 is a precedent-setter, not a jurisdiction-specific concern.
The question of when AI "provides" a licensed professional service — as opposed to assisting a licensed professional — is the central unresolved question in every jurisdiction where professional licensing exists. New Hampshire is the first state to convert that question into statute. Other states are watching.
For professional services firm owners with multi-state practices: the regulatory direction is set. The "meaningful oversight" standard will appear in other state bills, bar association guidance, accounting board rules, and eventually federal guidance. The firms that figure out what meaningful oversight actually looks like in their practice — not just in policy documents but in their actual workflow — will have a durable compliance answer regardless of which state adopts the next version of this rule.
The Two-Step That Works
For any firm with New Hampshire clients or that provides services in New Hampshire:
Step 1: Map where AI generates client-facing output.
List every workflow where AI creates something that goes to a client: AI-drafted advice, AI-prepared documents, AI-generated analysis, automated responses to client inquiries. Each one of these is a point where the "meaningful oversight" question can be raised.
Step 2: Attach documented human review to each touchpoint.
For each client-facing AI output, document that a named licensed professional reviewed the specific output before delivery. Not "a licensed professional is responsible for all AI output in this firm." That a specific licensed person reviewed this specific document before this client received it.
This is the standard that survives scrutiny. It's also good practice under ABA Formal Opinion 512, the AICPA's AI guidance, and every professional responsibility framework currently in use. New Hampshire SB 640 just made it a statutory requirement for firms serving New Hampshire clients.
What to Watch For
NH SB 640 is moving through the 2026 legislative session. Watch for:
- Final passage: The bill passed committee in March 2026. Full floor vote and governor signature are the remaining steps.
- Definition amendments: The final bill language may define "meaningful oversight" more specifically, or it may leave that to regulatory guidance. Any definition will be significant.
- Companion state bills: New Hampshire is likely to be followed by other states adopting similar "meaningful oversight" standards for professional services AI. California, New York, and Massachusetts are the most likely next movers.
The Troutman Privacy Blog publishes weekly state AI law updates and is the most reliable source for tracking NH SB 640's status and companion bills. Subscribe to that or add it to your firm's legal news monitoring.
Related Reading
Frequently Asked Questions
What does New Hampshire SB 640 say?
NH SB 640, which passed committee the week of March 9, 2026, states that AI systems may not 'provide licensed professional services' without a licensed professional's meaningful oversight. It targets law, accounting, medicine, and related professions. The bill is moving toward full legislative passage in New Hampshire's 2026 session.
What does 'meaningful oversight' mean under NH SB 640?
The bill does not define 'meaningful oversight.' This is the key ambiguity for professional services firms. The standard will likely be defined by enforcement actions or litigation — meaning the first cases or complaints will shape what counts. In the absence of a definition, the conservative compliance posture is: any AI-generated client output should be reviewed by a licensed professional before delivery, and that review should be documented.
Does NH SB 640 apply to firms outside New Hampshire?
The bill's scope as written targets AI use in providing professional services in New Hampshire. Firms with New Hampshire-based clients, or licensed professionals providing services to New Hampshire residents, may fall within scope. Multi-state professional services firms should monitor the bill's final language for jurisdictional reach provisions.
How is NH SB 640 different from Oregon's chatbot law?
Oregon HB 4154 targets AI-as-human deception in consumer-facing interactions (disclosure + private right of action). NH SB 640 targets the specific question of AI providing licensed professional services — the unauthorized practice angle. Oregon says 'disclose that it's AI'; New Hampshire says 'a licensed human must meaningfully oversee any AI professional services output.' They address different failure modes in the same general direction.
What is the first state to regulate AI in professional services this specifically?
NH SB 640 is among the first bills anywhere to directly address the question of when AI 'provides' a licensed professional service — the line between AI assistance and AI practice. Previous state laws have focused on discrimination, transparency, or general AI risk. NH SB 640 targets the professional licensing boundary specifically.