New Jersey Starts Asking Investment Advisers What Their AI Is Actually Doing With Client Money

There is no New Jersey AI statute for investment advisers. There is no rule proposal, no notice-and-comment docket, no press conference about frontier models. What there is, instead, is a questionnaire — and roughly 800 firms that had until June 30 to answer it.

On June 8, 2026, New Jersey Attorney General Jennifer Davenport and the Bureau of Securities, housed in the state's Division of Consumer Affairs, announced that the Bureau's annual investment adviser examination was underway. This year's edition asks state-registered firms to describe, in writing, how they use artificial intelligence in constructing investment portfolios, models, or recommendations to clients — including whether AI is used merely as a data-gathering or research tool — and how they market, advertise, or represent that use to the people whose money they manage.

It is a small instrument aimed at a narrow population. It is also a clear example of the pattern that will shape AI compliance far more than anything stuck in Congress: sectoral regulators applying decades-old fiduciary and anti-fraud law to machine-generated advice, without waiting for an AI statute to exist.

What New Jersey actually did

Precision matters here, because language around this kind of action gets sloppy fast. The Bureau did not open a formal enforcement sweep, issue targeted sweep letters, or propose a rule. It updated the questions in its recurring annual written examination — a compliance risk-assessment instrument every New Jersey-registered adviser must complete electronically each year.

The Bureau describes the examination as a risk assessment tool covering the nearly 800 New Jersey-registered investment adviser firms managing state residents' accounts. The 2026 version asks four things of substance: whether the firm uses AI in any manner to generate portfolios, models, or client recommendations; how the firm markets or discloses that use; whether it maintains written cybersecurity policies and conducts ongoing periodic staff training on them; and whether it performs periodic due diligence on vendors, including vendors' cybersecurity practices.

Responses were due June 30, 2026. Failure to comply, the Bureau warned, "may be deemed a failure to cooperate with the Bureau in violation of N.J.A.C. 13:47A-14.16, and result in administrative action." That is the teeth: not an AI rule, but an existing cooperation regulation.

Nor is this New Jersey's first pass. The Bureau's 2024 annual examination also focused on AI, and its 2019 edition centered on cybersecurity — a lineage the law firm Bressler, Amery & Ross noted in a June 22 client alert, calling both topics confirmed Bureau priorities rather than a one-year novelty.

The quotes, and what they signal

"Artificial intelligence and other emerging technologies may create new opportunities in the investment marketplace, but they also present new risks that demand careful oversight," said Bureau Chief Keith A. Alt. "By examining how firms use and market AI, as well as the steps they take to safeguard client data and manage cybersecurity threats, we can better identify emerging issues and promote transparency across the securities industry."

Note the verb: market. The Bureau is not primarily asking whether AI models are accurate. It is asking whether firms are telling the truth about them.

Jeremy E. Hollander, acting director of the Division of Consumer Affairs, was blunter about the mechanism: "The information gleaned from this examination enables our Bureau of Securities to identify emerging risks and address potentially disruptive industry practices before they can harm investors." In other words, the questionnaire is reconnaissance. What firms disclose this summer builds the map for whatever enforcement follows.

Analysis: the quiet path is the fast path

Washington has spent 2026 fighting over whether states should be allowed to regulate AI at all. That fight is largely irrelevant to what happened in Trenton, because New Jersey isn't regulating AI. It's regulating investment advisers — an authority nobody disputes — and simply pointing the existing questions at a new technology.

The federal template is already built. On March 18, 2024, the SEC announced settled charges against two advisers, Delphia (USA) Inc. and Global Predictions Inc., for false and misleading statements about their AI use, coining the enforcement shorthand "AI washing." Delphia paid $225,000; Global Predictions paid $175,000. The charges rested on Sections 206(2) and 206(4) of the Investment Advisers Act of 1940 plus the Marketing Rule and Compliance Rule. Nothing in that case required an AI-specific law. Every public statement an adviser makes must be fair, balanced, and substantiated — a standard that predates transformers by 80 years and applies cleanly to a firm that claims its models do more than they do.

State regulators have explicitly claimed this lane. On December 5, 2025, NASAA wrote to Congress opposing broad federal preemption of state AI laws, arguing states are better positioned to move quickly against AI-enabled fraud and citing FBI data on $50.5 billion in losses tied to AI-facilitated schemes. FINRA's 2026 annual regulatory oversight report added new content on generative AI and cyber-enabled fraud, and AI features in the SEC's FY2026 examination priorities alongside fiduciary conduct and compliance program effectiveness.

Stack those and an adviser using AI in 2026 faces layered scrutiny from a state bureau, a federal commission, and a self-regulatory organization — none of which needs new legislative authority to ask hard questions. The binding constraint on AI in financial services will not be an AI act. It will be Section 206, the Marketing Rule, and a state questionnaire with a June 30 deadline.

What to watch

Three things. First, whether the Bureau publishes aggregate findings from the 2026 responses — the first real state-level census of how smaller advisers actually deploy AI. Second, whether New Jersey follows up with targeted exams or an enforcement action grounded in the questionnaire's disclosures, the step that would convert reconnaissance into a genuine sweep. Third, whether other state securities administrators copy the question set. NASAA has already aligned state advertising rules with the SEC Marketing Rule; a shared AI-disclosure module would spread New Jersey's approach across dozens of jurisdictions without a single new statute passing anywhere.