What U.S. State Consumer Protection Statutes and Their Enforcement Histories Actually Say About AI-Generated Legal Documents Sold Directly to Consumers — and Which State Attorneys General Have Moved First

The Legal Stack | Regulatory / Access to Justice | October 24, 2026

The market for AI-generated legal documents sold directly to consumers has grown rapidly alongside the broader expansion of legaltech. Tools that draft wills, leases, demand letters, and immigration forms without any lawyer in the loop are now widely available, and at least some of them make claims — explicit or implied — about the legal adequacy of what they produce. Regulators are beginning to respond, though enforcement remains uneven and doctrinal frameworks are still being assembled. This briefing synthesises what the public record actually shows across four domains: FTC guidance, state bar UPL opinions, state AG enforcement, and the access-to-justice counterargument that legaltech companies advance in their defence.

1. The FTC's Posture on AI Products and Implied Legal Adequacy Claims

The Federal Trade Commission has not issued a rule or formal guidance document specifically addressing AI-generated legal documents. What it has done is apply its long-standing authority under Section 5 of the FTC Act — which prohibits unfair or deceptive acts or practices in commerce — to AI product marketing more broadly, and one enforcement action is directly relevant.

In September 2024, the FTC filed a complaint against DoNotPay, the company that marketed itself as "the world's first robot lawyer." The proposed consent order, finalized in January 2025 and publicly announced in February 2025, required DoNotPay to pay $193,000 in monetary relief and to notify subscribers from 2021 through 2023 of the action. The order also prohibited the company from advertising AI lawyer functionality without substantiating evidence that its outputs perform comparably to a licensed attorney. The FTC's theory of liability was deception: the agency alleged that DoNotPay's "robot lawyer" marketing conveyed an implied claim of legal competence that the product could not substantiate.

The DoNotPay matter is the closest the FTC has come to a doctrinal statement on AI legal document adequacy. Several analytical threads emerge from it. First, the FTC treated implied performance claims — not just express warranties — as actionable misrepresentations. A tool that markets itself as capable of producing a legally enforceable contract, or of protecting a consumer's rights in a specific proceeding, may be making a substantiation-required claim even if the marketing copy never uses the word "guarantee." Second, the FTC's monetary relief figure was modest, reflecting the agency's preference in novel enforcement contexts to establish precedent before escalating penalties. Third, and significantly, the FTC's action predated and in some respects prompted the wave of state-level activity that followed.

The FTC has separately signalled, through blog posts and public statements from its Bureau of Consumer Protection, that AI-specific claims about reliability, accuracy, and professional equivalence are subject to the same substantiation standards as any other performance claim. Sellers of AI legal document tools who cannot demonstrate that their outputs meet the legal standard of care applicable to the specific legal matter at hand — a standard that varies by jurisdiction, document type, and individual circumstance — face potential exposure under Section 5.

2. State Bar UPL Opinions on AI Document Generation Since 2024

California. The State Bar of California published its Practical Guidance on Generative AI in November 2023, addressed primarily to licensed attorneys. It did not directly opine on consumer-facing AI legal document services. The California Committee on Professional Responsibility and Conduct approved proposed AI rule amendments to the California Rules of Professional Conduct in March 2026, which were in public comment through May 2026 and had not been formally adopted as of October 2026. The State Bar has not issued a formal unauthorized practice of law opinion specifically addressing AI document generation sold to consumers, and California's UPL statute — Business and Professions Code Section 6125 — does not contain AI-specific provisions. California's enforcement of UPL against document preparation services has historically proceeded through the State Bar's intake and investigation process, but no publicly documented enforcement action specifically targeting an AI document generation company has been identified in the public record.

New York. The New York City Bar Association issued Formal Opinion 2024-5 on August 7, 2024, addressing lawyers' use of generative AI, and Formal Opinion 2025-6 in December 2025, addressing AI recording and transcription. Neither opinion addresses the UPL question directly in the context of consumer-facing tools. The New York State Bar Association has published guidance on generative AI for lawyers but has not issued a formal opinion specifically classifying AI-generated consumer legal documents as the practice of law or as UPL. The doctrinal question — whether an AI tool that produces a completed will or lease without any human lawyer in the loop constitutes the practice of law under New York Judiciary Law Section 478 — remains formally unresolved, though New York courts have historically construed UPL broadly.

Texas. The Texas Professional Ethics Committee issued Opinion 705 in February 2025, addressing lawyers' use of generative AI in practice. Like its California and New York counterparts, the opinion is addressed to licensed practitioners and does not directly address whether consumer-facing AI document generation constitutes UPL under the Texas Disciplinary Rules or the Texas Government Code's UPL provisions. No formal Texas bar opinion specifically classifying AI document generation as UPL has been identified in publicly available records.

The net result across all three jurisdictions is the same: state bars have been active in addressing AI for lawyers, but have not yet issued opinions directly determining whether AI-generated consumer legal documents constitute unauthorized practice. This leaves a doctrinal gap that state attorneys general, rather than bar disciplinary authorities, have begun to fill — imperfectly and through the different lever of consumer protection law.

3. State AG Consumer Protection Enforcement: What Has Actually Happened

The FTC's DoNotPay action remains the clearest federal precedent, but it was the FTC — not a state AG — that brought it. As of October 2026, no state attorney general has filed a publicly documented consumer protection enforcement action specifically targeting an AI-generated legal document service. This is not an empty observation. Several state attorneys general — including those of California, Illinois, and New York — have publicly indicated interest in AI consumer protection generally, and California's Consumer Protection and Economic Justice section has issued guidance on AI product marketing. However, converting general interest into a targeted enforcement action against a legaltech company requires establishing concrete consumer harm, which in legal document contexts is methodologically difficult: a defective will may not cause demonstrable harm until the testator is dead, and a defective lease may never be tested if the landlord never litigates.

The California Privacy Protection Agency did bring an enforcement action against Honda in March 2025 ($632,500) and a separate AG settlement against DoorDash in February 2024 ($375,000), demonstrating active enforcement against technology companies on consumer protection grounds — but neither involved legal document AI. These cases are relevant context for the tools available to state regulators, not precedents on the substantive question.

What has happened, rather than formal enforcement, is a pattern of regulatory signalling. Several state AGs have issued consumer alert guidance warning consumers about AI legal tools, and at least one legislative proposal — in Illinois, building on the state's existing regulatory posture under the Human Rights Act amendment effective January 1, 2026 — has sought to require AI legal document tools to disclose their non-attorney status at the point of sale and to specify the limitations of their outputs. Whether that proposal advances to enactment is not yet clear from the public record.

4. The Doctrinal Tension: Access to Justice vs. Consumer Protection

Legaltech companies offering direct-to-consumer AI legal documents have a powerful access-to-justice argument available to them, and they use it. The 2022 Legal Services Corporation Justice Gap Report, conducted by NORC at the University of Chicago, found that 92 percent of low-income Americans' civil legal problems received no or insufficient professional help, and that 74 percent of low-income households experienced at least one civil legal problem in the prior year. LSC-funded organizations turned away 49 percent of requests due to limited resources. These figures represent a structural failure of the existing attorney-centred delivery model, and they give direct-to-consumer legal document tools a genuine public interest framing.

State regulators are not insensible to this argument. Several state bar UPL committees have acknowledged, in guidance documents addressed to their own members, that rigid UPL enforcement against non-attorney legal services can itself harm the consumers those rules purport to protect. Utah's legal regulatory sandbox, approved in August 2020, is the most concrete institutional recognition of this tension — it permits non-lawyer-owned entities to provide legal services under regulatory oversight, and some AI legal document companies have sought to operate within it.

The consumer protection counterargument, however, is not about restricting access to cheap legal help. It is about ensuring that what consumers receive when they pay for an AI-generated legal document is what they believe they are purchasing. If a consumer buys an AI-generated will believing it will be enforceable in their state, and the will fails attestation requirements that vary from state to state, the access-to-justice framing does not resolve the consumer protection problem — it potentially compounds it, because the consumer believed they had addressed a legal need when they had not.

This is the doctrinal tension that neither the FTC's DoNotPay action nor the state bars' AI guidance has fully resolved. The FTC established that implied competence claims are actionable. The state bars have not yet drawn the line between permissible document automation and UPL in the AI context. And no state AG has yet brought a case that would force a court to work through the analysis. The enforcement landscape, in other words, is one of significant potential liability, limited precedent, and a regulatory race — between federal and state authority, and between consumer protection and access-to-justice frameworks — that has not yet produced a winner.

Research sourcing: FTC complaint and consent order, DoNotPay (September 2024 / February 2025); ABA Formal Opinion 512 (July 2024); NYC Bar Formal Opinions 2024-5 and 2025-6; Texas Professional Ethics Committee Opinion 705 (February 2025); California State Bar Practical Guidance on Generative AI (November 2023); LSC 2022 Justice Gap Report (NORC at U. Chicago); Utah legal regulatory sandbox (approved August 2020); California CPPA enforcement record (public); Colorado AI regulatory history (public legislative record). All enforcement characterisations reflect publicly documented actions as of October 2026.