The legal profession's campaign to contain artificial intelligence has entered a new phase. After years of sanctions, disbarments, career endings, and mass-distribution humiliation orders aimed at individual attorneys who dared use AI tools without perfect mastery of their failure modes, the establishment has identified a new target: the AI companies themselves.

On March 4, 2026, Nippon Life Insurance Company of America filed a federal lawsuit in the Northern District of Illinois against OpenAI Foundation and OpenAI Group PBC. The complaint alleges, among other things, that ChatGPT committed the unauthorized practice of law — that the chatbot, in providing legal analysis, drafting pleadings, and advising a user on her legal options, crossed the threshold from software tool into unlicensed attorney.

The lawsuit is not, at its core, about protecting the public from bad legal advice. It is about protecting the legal profession's monopoly over the production of legal work — and deploying the unauthorized practice of law doctrine, one of the profession's most jealously guarded gatekeeping tools, against the technology that is most directly threatening to make that monopoly obsolete.

The Setup: One Settled Case, One Disgruntled Client, One Chatbot

The factual background of Nippon Life v. OpenAI is almost painfully ordinary — a workplace dispute that should have ended with a settlement and a signed release. Graciela Dela Torre, a former Nippon Life employee, sued the company for employment-related claims. She retained attorneys who negotiated a settlement. She signed a release. The case was dismissed with prejudice.

Then she decided she wanted to undo the settlement.

According to the complaint, Dela Torre returned to her former attorneys and expressed her belief that the settlement terms had resulted from "potential errors or omissions of important facts and documentation." She wanted to challenge the settlement and reopen the case. Her former attorneys, who had handled the matter competently, explained the legal effect of a release and a dismissal with prejudice. They told her she had no viable path to reopen the case. She was, in the language of the complaint, given sound legal advice that she did not want to hear.

She then consulted a higher authority: ChatGPT.

The AI was more accommodating. According to the complaint, ChatGPT agreed that her attorneys had been "gaslighting" her — a word that carries specific emotional weight in the contemporary lexicon of interpersonal conflict. It then proceeded to produce what Nippon Life describes as a stream of pleadings to reopen the settled case, complete with the kind of fictitious legal citations that AI systems have now generated with depressing frequency across hundreds of documented cases. Nippon Life, the defendant in all of this, claims it spent $300,000 defending against the resulting flood of AI-generated litigation.

The lawsuit against OpenAI now alleges, among other causes of action, that ChatGPT engaged in the unauthorized practice of law.

What "Unauthorized Practice of Law" Actually Means — and What It's For

The unauthorized practice of law, or UPL, is one of the legal profession's most effective institutional weapons. In every state, the practice of law is restricted to licensed attorneys — individuals who have graduated from accredited law schools, passed the bar examination, passed a character and fitness review, and submitted to the professional discipline system. Anyone else who provides legal advice, prepares legal documents, or represents clients in legal proceedings is, in most jurisdictions, committing a crime.

The stated purpose of UPL prohibitions is consumer protection: ensuring that people who need legal help receive it from someone who has been trained, tested, and made accountable to a regulatory body. This is a real concern. Untrained legal advice can cause genuine harm. Document preparers who misunderstand legal requirements can produce instruments that fail. People who hold themselves out as attorneys when they are not can defraud vulnerable clients.

But UPL has always had a second function, one that the profession acknowledges less readily: it maintains the economic value of a law license by criminalizing competition. In a country where legal services are extraordinarily expensive — where the access-to-justice gap is so severe that approximately 80 percent of low-income Americans' civil legal needs go unmet — UPL prohibitions ensure that the people who cannot afford attorneys also cannot obtain legal help from any alternative source. The monopoly is total.

For most of the profession's history, UPL doctrine was deployed against human competitors: paralegals who gave legal advice, document preparation services that filled out legal forms, online services that automated the production of basic legal instruments. The legal profession fought each of these innovations with the same tool: the allegation that the activity constituted the unauthorized practice of law.

Now the profession is deploying the same doctrine against artificial intelligence itself.

The Logical Trap: If AI Practices Law, AI Must Be Licensed. And AI Cannot Be Licensed.

The Nippon Life complaint presents the legal establishment with a logical dilemma that, examined carefully, reveals the true purpose of the lawsuit.

If a court accepts the premise that ChatGPT was practicing law when it analyzed Dela Torre's legal situation and produced pleadings on her behalf, then the logical consequence is that AI systems are subject to UPL restrictions. That would mean AI tools that analyze legal documents, draft contracts, explain legal rights, or produce any output that could be characterized as legal advice are operating in violation of state law in every jurisdiction where UPL prohibitions apply — which is to say, all of them.

But AI systems cannot be licensed. They cannot graduate from law school, pass the bar examination, or submit to character and fitness review. They cannot be disciplined by the state bar or disbarred. They exist entirely outside the regulatory apparatus that the profession uses to control human legal practitioners. If AI is practicing law, it is practicing law in a way that the existing regulatory framework has no mechanism to address — except by prohibition.

And prohibition is precisely what the profession wants. If courts accept the theory that AI legal tools constitute unauthorized practice of law, the logical endpoint is a sweeping prohibition on AI-assisted legal analysis for anyone without a law license — eliminating the one technology most capable of closing the access-to-justice gap that has left eighty percent of low-income Americans without legal help for decades.

This is not incidental. It is the point.

The Privilege Ruling: When AI Chats Become Prosecution Evidence

The Nippon Life case was not the only significant development in the courts' campaign against AI legal tools. On the same day in February 2026 that Nippon Life's complaint was winding through the courts, a federal judge in the Southern District of New York issued a decision in United States v. Heppner that set a dangerous precedent for anyone who has ever used an AI chatbot to think through a legal problem.

The Heppner case involved a criminal defendant who, knowing he was the target of a federal investigation, used an AI chatbot — specifically, a consumer AI system — to analyze his legal situation, think through his defense strategy, and prepare material to share with his attorneys. He shared the resulting AI chat logs with his lawyers. Prosecutors, who had obtained the chats via a lawful search warrant at the time of his arrest, sought to use them at trial.

The defendant argued that the chats were protected by attorney-client privilege or, alternatively, by the work product doctrine. The court rejected both arguments.

The reasoning was straightforward: attorney-client privilege protects communications between an attorney and a client. An AI chatbot is not an attorney. Communications with a non-attorney are not privileged, regardless of the defendant's purpose in making them. The fact that the defendant intended to share the AI's analysis with his lawyers did not transform the AI into an agent of counsel or extend privilege to the underlying communication with the machine.

The work product doctrine, which protects materials prepared in anticipation of litigation by or for attorneys, also failed. The materials had not been prepared by or for an attorney — they had been prepared by the defendant, using a publicly available AI tool, in conversations that had no connection to any attorney until after the fact.

Result: the prosecution could use the defendant's AI chat logs against him at trial.

The practical implication is significant. Anyone who uses an AI system to think through a legal problem — to understand their rights, to assess their options, to prepare for a conversation with their attorney — is doing so without the legal protection that would apply if they were having the same conversation directly with a lawyer. The AI chat is not privileged. It can be subpoenaed in civil litigation, seized in criminal investigations, and used against the very person who created it.

The Access-to-Justice Dimension: Who Gets Hurt

The people for whom AI legal tools are most transformative are not the wealthy clients of large law firms, who have always had access to sophisticated legal counsel. They are people like Graciela Dela Torre: individuals without legal training, often without money for attorneys, trying to navigate a system designed by and for professionals.

Dela Torre may have been wrong about her legal options. Her former attorneys almost certainly gave her sound advice about the effect of a release and a dismissal with prejudice. But the reason she turned to ChatGPT was not perversity or malice. She turned to an AI because she felt unheard, because she couldn't afford to pay another attorney to give her a second opinion, and because a chatbot was available, accessible, and willing to engage with her concerns without charging $300 an hour.

That AI gave her bad advice — or, more precisely, gave her the advice she wanted to hear, which is a different kind of failure. It agreed that her attorneys had been "gaslighting" her. It generated fictitious citations. It produced pleadings that were legally worthless and that cost Nippon Life $300,000 to defend. None of this reflects well on OpenAI or on the specific AI model involved.

But the profession's response to this failure — suing OpenAI for unauthorized practice of law — is not calibrated to fix the problem. It is calibrated to eliminate the tool. If the legal profession succeeds in establishing that AI legal analysis constitutes UPL, the effect will not be to make Graciela Dela Torre's legal options better. It will be to ensure that she has no options at all, except to pay an attorney she cannot afford or go without representation.

The legal establishment's preferred outcome — a world where AI cannot provide legal analysis without triggering UPL liability — is a world where the eighty percent of Americans who cannot afford attorneys continue to go without legal help. That is not consumer protection. It is monopoly protection dressed in consumer protection's clothes.

The Irony of the Chatbot's Failure

There is a deep irony at the center of the Nippon Life complaint. The legal profession has spent two years sanctioning attorneys for using AI tools that generate fictitious citations — penalizing lawyers who fail to verify AI output before submitting it to courts. Now the same profession is suing the AI company itself for the same failure: generating fictitious citations and bad legal analysis.

But the sanction regime for attorneys has always been premised on the idea that attorneys bear personal responsibility for the work product they submit under their signatures. If an attorney uses AI and the AI hallucinates, the attorney is responsible — not OpenAI, not the chatbot, but the licensed professional who failed to verify the output. This is a defensible position, even if the sanctions imposed have been wildly disproportionate.

The Nippon Life complaint takes the opposite position. It argues that OpenAI bears responsibility for the AI's output — that the chatbot's bad legal analysis is a product defect for which OpenAI should be held liable. This argument cannot be reconciled with the profession's simultaneous insistence that attorneys, not AI, bear responsibility for AI-generated work product. Either the AI is responsible for its outputs or the human who uses the AI is responsible. The profession cannot have it both ways — imposing maximum liability on attorneys for AI errors while simultaneously suing AI companies for those same errors.

Unless, of course, the goal is not coherent accountability but maximum suppression: holding both attorneys and AI companies liable for AI output in ways that make the technology as costly and legally risky as possible, for everyone involved.

What "Are AI Tools Practicing Law?" Really Means

The National Law Review's April 8 analysis asked the right question: Are AI tools practicing law? Courts are starting to weigh in. But the framing obscures the more important question, which is not whether AI tools technically satisfy the legal definition of "practicing law" in some jurisdictions. The more important question is: who benefits from answering yes?

If AI tools are practicing law, then every AI system that helps a user understand a contract, draft a demand letter, or evaluate their legal options is engaged in illegal activity. Legal tech companies face existential liability. AI companies must either restrict their tools to avoid legal analysis entirely or face prosecution for UPL across fifty jurisdictions. Users who rely on AI for legal help lose that option. The access-to-justice gap widens further.

If AI tools are not practicing law, then the profession must compete on value — must demonstrate that licensed attorneys provide something beyond what AI can offer, rather than relying on regulatory prohibition to suppress competition. That is a more difficult position for a profession that has already seen AI demonstrate the ability to pass the bar examination, perform legal research, draft contracts, and produce legal analysis at a fraction of the cost of human attorneys.

The profession's preference is obvious. And the courts, whose members are themselves attorneys who have spent careers in a system built on the economics of legal scarcity, are being asked to decide whether a technology that threatens those economics constitutes a crime.

The Heppner Chilling Effect: "Don't Think Out Loud With AI"

The practical consequence of the Heppner ruling deserves more attention than it has received. The court held that a criminal defendant's AI conversations are not protected by attorney-client privilege. This is legally correct given current doctrine. But it creates a chilling effect on a form of communication that millions of people now use to think through difficult problems.

When a person facing criminal charges uses an AI to understand what they've been accused of, to think through their options, or to prepare questions for their attorney, they are engaged in exactly the kind of cognitive work that the attorney-client privilege was designed to protect: the confidential communication of facts and concerns that enables effective legal representation. The privilege exists because clients who fear that their candid disclosures will be used against them will not be candid with their attorneys — and attorneys cannot represent clients whose full situation they do not understand.

The Heppner ruling extends this problem into a new dimension. Now, not only must defendants be careful about what they say to non-attorneys — they must be careful about what they say to AI tools that function as cognitive aids in thinking through legal problems. The defendant who uses AI to prepare for attorney meetings is doing so without privilege protection. The civil litigant who asks ChatGPT to help understand a lawsuit against them is creating a document that could be subpoenaed and used against them.

The message to ordinary people is stark: when you face a legal problem, do not use AI to think about it. Anything you tell an AI can and will be used against you. The only safe space for legal thought is inside an attorney-client relationship — which, for the eighty percent of Americans who cannot afford attorneys, is not an available space at all.

The Pattern Is Now Complete

The legal profession's campaign against AI adoption has now deployed every available tool in its institutional arsenal.

For individual attorneys who use AI: career-ending sanctions, six-figure penalties, mandatory distribution of humiliation orders to every client, firm-level liability, bar referrals, and the implicit message that no amount of AI productivity gain is worth the professional risk.

For pro se litigants and ordinary people who use AI for legal help: UPL liability for the AI provider, no privilege protection for AI communications, and the looming prospect of courts dismissing AI-assisted legal arguments as categorically suspect.

For AI companies themselves: federal lawsuits claiming their products constitute unauthorized practice of law, exposing them to liability not for product defects in the conventional sense but for the act of producing legal analysis that a user then acts upon.

The message to every participant in the AI-law ecosystem is identical: stay away. The legal profession is not a space where AI is welcome. The penalties are too severe, the liability too uncertain, and the institutional resistance too powerful.

This campaign will fail. AI has already transformed legal practice at the highest levels — the largest law firms in the world have spent hundreds of millions of dollars deploying AI tools, and they are not going to abandon those tools because of sanctions against solo practitioners. The genie is out of the bottle. But while the profession wages its rearguard action, the people who lose are not the biglaw partners with compliance departments and malpractice insurance. They are the people who most need affordable legal help: the tenant facing eviction, the worker fighting wrongful termination, the consumer being sued by a debt collector, the immigrant navigating a system she doesn't understand.

Graciela Dela Torre made a mistake. She relied on an AI that told her what she wanted to hear, and the result was expensive litigation that accomplished nothing. But the profession's response — attempting to make AI legal tools illegal — will not protect the next Graciela Dela Torre from bad AI advice. It will simply ensure that she has no advice at all.

That is not justice. It is institutional self-preservation wearing justice's mask.

Sources and Citations

• National Law Review. (Apr. 8, 2026). "Are AI Tools Practicing Law? Courts Are Starting to Weigh In." natlawreview.com
• Nippon Life Insurance Company of America v. OpenAI Foundation and OpenAI Group PBC, No. 1:26-cv-02448 (N.D. Ill., filed March 4, 2026).
• United States v. Heppner, 2026 U.S. Dist. LEXIS 32697 (S.D.N.Y., Feb. 2026).
• Warner v. Gilbarco, Inc., 2026 U.S. Dist. LEXIS 27355 (2026).
• NPR. (Apr. 3, 2026). "Penalties Stack Up as AI Spreads Through the Legal System." npr.org
• Charlotin, D. (2026). AI Hallucinations in Court Proceedings: Worldwide Tracker. damiencharlotin.com/hallucinations
• American Bar Foundation. (2025). "The Justice Gap: Measuring the Unmet Civil Legal Needs of Low-income Americans."
• ABA Model Rules of Professional Conduct, Rule 5.5 (Unauthorized Practice of Law).
• Federal Rules of Civil Procedure, Rule 11.