FTC Final Order Against DoNotPay: Legal Risks of "AI Lawyer" Claims and Performance Substantiation

On February 11, 2025, the U.S. Federal Trade Commission (FTC) announced that it had finalized a consent order (Decision and Order) against DoNotPay, Inc., which had promoted its service as "the world's first robot lawyer." The order requires payment of USD 193,000 in monetary relief, notification of existing customers from the relevant period, and prohibitions concerning future claims. The case was brought as part of "Operation AI Comply," a sweep announced by the FTC in September 2024, and was finalized after a public comment period. This article confirms from primary sources the legal nature of the order, the claims the FTC took issue with and why they lacked substantiation, and the specific obligations the order imposes, and then organizes the issues under the Premiums and Representations Act and the Attorney Act that arise when Japanese companies make "AI lawyer" claims and the like, as a framework separate from the U.S. order.

Background to the Order and Its Legal Nature

On September 25, 2024, the FTC announced its allegations and proposed consent agreement as part of a sweep against several businesses, including DoNotPay. The Commission voted 5–0 to approve issuance of the complaint and proposed consent agreement, which then entered a 30-day public comment period. The FTC explains that five comments were received and that the Commission voted 5–0 on January 16, 2025 to approve the final order. The Complaint and the Decision and Order issued by the Commission are both dated January 14, 2025, and the docket number is Docket No. C-4812.

This order is not a federal district court judgment but an administrative consent order under Section 5 of the FTC Act. The decision expressly states that DoNotPay neither admits nor denies the facts alleged in the complaint, and admits only the facts necessary to establish jurisdiction for the purposes of this proceeding; it does not legally admit the facts regarding the substance of the claims themselves. However, the facts alleged in the complaint will be deemed true without proof in certain future enforcement proceedings, such as proceedings to except the obligation from discharge in bankruptcy if DoNotPay fails to perform its monetary payment obligation. The addressees of the order are limited to DoNotPay, Inc. and its successors and assigns, and the order is understood not to directly bind other AI legal service businesses.

The Claims the FTC Took Issue With and Why They Lacked Substantiation

According to the complaint, DoNotPay promoted itself as "the world's first robot lawyer" on its website, in app stores and in advertising on TikTok and elsewhere, advertising that users could file lawsuits at the push of a button, and its CEO said that it would "replace the USD 200 billion legal industry with artificial intelligence." The company's site displayed a quotation stating that what this robot lawyer could do was surprisingly similar to, or even better than, what a human lawyer does, but the complaint points out that the quotation actually came from an opinion piece by a single high school student published on a blog for high school students run by the Los Angeles Times.

The central claims the complaint took issue with were claims that the service, like a human lawyer, could apply the law to a client's specific facts, avoid pitfalls such as statutes of limitations, damages caps and jurisdiction based on expertise, and detect and advise on legal violations on businesses' websites. In addition, the claims for a "website check" function, which supposedly could detect hundreds of federal and state law violations using only an email address and prevent a situation that would otherwise require USD 125,000 in attorney's fees if left unaddressed, are also said to have been made when the service did not actually perform such analysis.

The reasons the FTC found the claims unsubstantiated are clear. DoNotPay never tested whether the output of its law-related features was at the level of a human lawyer; the service combined natural language processing models, a chatbot and OpenAI's ChatGPT API, and was not trained on a comprehensive and current corpus of federal and state law and case law. The company neither employed nor retained attorneys to verify quality and accuracy. Separately, the State Bar of California had been investigating the company since 2021 on suspicion of the unauthorized practice of law, but the concurring statement of Commissioner Ferguson expressly states that the FTC's action in this case should not be equated with enforcement of state regulations on the unauthorized practice of law, and that the Commission has no authority to enforce state professional licensing laws. This is because regulation based on lack of substantiation for claims and regulation prohibiting the handling of legal matters by non-lawyers as such differ in legal nature.

Specific Obligations Imposed by the Order

The Decision and Order requires DoNotPay to pay USD 193,000 to the FTC within 30 days after the order becomes effective. In addition, it requires email notification, using specified text, within 180 days after the order is issued, to customers who subscribed between January 1, 2021 and December 31, 2023. The notification informs consumers that the FTC found some of the company's advertising to be insufficiently substantiated, and that the company will not in future advertise that it can draft documents, file lawsuits or give legal advice like a human lawyer unless it has sufficient substantiation.

The scope of the restrictions on future claims is not limited to legal services. The "covered products or services" defined in the order include, in addition to the DoNotPay service, any internet product or service offered by the company that purports to provide "professional services" such as legal, accounting, architectural, consulting or medical services. The order prohibits claims that these function like a human professional or have professional-level performance unless, at the time the claim is made, the company possesses "competent and reliable evidence" that has been objectively evaluated by qualified persons and is generally accepted in the profession to yield accurate and reliable results. In addition, the order sets out a 10-year recordkeeping obligation, an obligation to submit compliance reports after one year and annually for three years thereafter, and a duration of 20 years from the date of issuance.

Points of Contact for Japanese Companies (Comparison with the Premiums and Representations Act and Article 72 of the Attorney Act)

Where a Japanese company makes similar "AI lawyer" claims, I think it is necessary to distinguish and consider at least two different legal issues.

The first is the issue of representations misleading as to quality under Article 5, item 1 of the Premiums and Representations Act. Representations that mislead consumers into believing that the performance of an AI service is significantly better than it actually is may become subject to investigation by the Consumer Affairs Agency. Under the regulation of unsubstantiated advertising in Article 7, paragraph 2 of the same Act, where the Consumer Affairs Agency requests substantiation, the business must in principle submit materials within 15 days, and if it cannot, the representation is deemed misleading. While the FTC order requires possession of substantiation at the time the claim is made, the regulation of unsubstantiated advertising is structured so that substantiation is submitted after the fact upon request; that point differs, but I think the two have in common that performance claims require supporting materials.

The second is the prohibition on the handling of legal matters by non-attorneys under Article 72 of the Attorney Act (Japanese). That article prohibits persons who are not attorneys or legal professional corporations from engaging, as a business and for the purpose of obtaining compensation, in providing legal opinions, representation, arbitration, settlement or other legal services with respect to lawsuits and other general legal cases. This regulation differs significantly from the FTC order in that it operates regardless of the truthfulness of the representation. Even if the performance claims are accurate, an issue under Article 72 may arise if the substance of the service is engaging in legal services amounting to legal opinions or representation in individual matters as a business for compensation; conversely, even if the performance claims are unsubstantiated, no issue under Article 72 arises if the substance stays within the provision of information or support for preparing forms. This division of roles is also taken up in the article organizing what an AI lawyer is.

Where a company uses claims such as "AI lawyer," "replaces lawyers" or "equivalent to a human professional" for its AI service, the starting point for a practical response is to check the following points by the next business day.

• An inventory of the claim wording. Identify expressions suggesting performance or equivalence with professionals, including on websites, in app stores and in social media posts.
• Substantiating materials for performance claims. Check whether supporting materials, such as third-party verification or test results by professionals, existed at the time the claim was made, and if not, consider revising the claim or preparing substantiating materials.
• Claims about the involvement of qualified professionals. Where the company makes claims suggesting that attorneys or qualified professionals supervise or are involved in the service, check whether the actual existence and extent of involvement match the content of the claims.

For reviewing terms of use and claims, The Legal Issues to Look at First in the Terms of Use of Generative AI Services and The First Legal Checklist a Company Starting an AI Service Should Create may also be helpful. One of the fundamental causes the FTC took issue with in the DoNotPay case was the absence of any system in which human professionals verified the AI's output. This point is an issue shared with the importance of a human performing the final check of generative AI answers.

Developments to Watch

The Decision and Order requires DoNotPay to submit a compliance report about one year after the date of issuance (around January 2026). As of the time of writing, neither the content of this report nor the FTC's evaluation of it has been made public. In addition, the FTC published a proposed policy statement on the accuracy of AI output in July 2026, but this is a separate, general policy-making development from this case, and this article does not go into it in depth. How far the approach shown in the DoNotPay order, under which claims purporting to provide professional services require substantiation at the time they are made, will be applied in future FTC enforcement against other AI legal services and AI services in the medical and accounting fields is a matter to keep monitoring.