Can AI Give Legal Advice? ChatGPT, UPL and Privilege (2027)

You can use ChatGPT or another AI chatbot to learn about the law, but it is not a lawyer: it cannot represent you, a federal court in New York ruled in February 2026 that one defendant's own chats with an AI platform were not protected by attorney-client privilege, you are responsible for anything you file in court, and the AI companies' own terms restrict tailored legal advice given without a licensed professional. As of November 6, 2026, we found no court ruling that an AI tool itself practices law; a pending federal lawsuit, Nippon Life v. OpenAI, raises that question.

Status as of November 6, 2026: OpenAI's motion to dismiss Nippon Life v. OpenAI (N.D. Ill.), which seeks dismissal of the whole complaint including its unlicensed-practice claim, is fully briefed and has not been ruled on; a status hearing is set for December 2, 2026. Revised usage policies from OpenAI (labelled effective November 20, 2026) and Anthropic (labelled effective November 12, 2026) are scheduled but not yet in effect.

This page covers U.S. law on unauthorized practice of law, privilege and court filings as it applies to people using AI chatbots, with California and New York as statutory examples. It is general legal information, not legal advice. For how courts regulate AI in filings and sanction fake citations, see our AI in Court guide.

Legal information vs. legal advice: what counts as legal advice?

There is no single national definition of legal advice. Each state decides what counts as the practice of law through its own statutes and court rules, and the American Bar Association's Model Rule 5.5(a) simply tells lawyers not to "practice law in a jurisdiction in violation of the regulation of the legal profession in that jurisdiction, or assist another in doing so." That Model Rule is addressed to lawyers; it does not define the practice of law for everyone else.

As a working line, legal information explains what the law says in general. Legal advice applies the law to one person's specific facts and tells that person what to do. A chatbot answering "what is a statute of limitations" is giving information; telling you which claims to file in your case, or drafting the papers for you to sign, moves toward advice.

Definition (one company's version): Anthropic's usage policy, in a version labelled effective November 12, 2026, describes the high-risk legal use case as "advising an individual on their legal rights, obligations, or liability, or on what to do in a specific legal matter, or preparing legal documents for them to sign or file." This is a company term, not a statute, but it tracks the common distinction.

State statutes show how the line is drawn in practice:

Other states use their own wording. Check your own state's statute or state bar before relying on a general rule.

Is it illegal for AI to give legal advice?

As of November 6, 2026, we found no court ruling, statute or state bar opinion holding that an AI tool itself practices law, or that a consumer commits unauthorized practice of law by asking a chatbot a legal question. That is a search-limited finding: it covers the statutes, court records, bar opinions and agency actions we opened for this page, not every court in the country.

Part of the reason is how the statutes are written. New York's Judiciary Law 478 and 484 speak of a "natural person," and California's B&P 6125 speaks of a "person." Whether software can fall within those words has not been decided in any ruling we found.

Nippon Life v. OpenAI: the pending unlicensed-practice claim

Nippon Life Insurance Co. of America v. OpenAI Foundation and OpenAI Group PBC, filed March 4, 2026 in the Northern District of Illinois (No. 1:26-cv-02448), is the live test. Nippon Life, an insurer previously sued by a disability claimant, alleges that ChatGPT helped that claimant pursue litigation against it. Its complaint pleads tortious interference with a contract, abuse of process, and unlicensed practice of law under Illinois's Attorney Act, 705 ILCS 205/1, alleging that "ChatGPT engaged in the practice of law in the state of Illinois without a license."

OpenAI moved to dismiss on May 15, 2026. Its brief argues that "ChatGPT is not a lawyer and it does not practice law," that "ChatGPT and its underlying LLMs are tools, not persons," and that only Illinois-licensed attorneys have standing to bring an unauthorized-practice claim. As of the docket read through October 8, 2026, the motion "remains under advisement," and a status hearing was reset to December 2, 2026. Everything above is an allegation or an argument; the court has made no finding.

Other actions that touch the question

• FTC and DoNotPay: In February 2025 the Federal Trade Commission finalized an order requiring DoNotPay, which marketed an "AI lawyer," to pay $193,000 and barring it from claiming its service performs like a real lawyer without sufficient evidence. That is a deceptive-advertising order, not a ruling on unauthorized practice of law.
• New York S7263A: A pending New York Senate bill would impose liability on a chatbot operator that knowingly lets its chatbot impersonate a licensed professional in a way that would be unauthorized practice if a natural person did it. The Senate's bill page shows it was committed to the Rules Committee on June 5, 2026. It is a bill, not law.
• Arizona: Arizona's Supreme Court licenses some nonlawyer-owned legal businesses as Alternative Business Structures; for example, Administrative Order 2026-47 (May 5, 2026) approved one. This shows states can authorize legal services outside the traditional law firm model. It does not mean any AI chatbot is licensed to give legal advice.

What the AI companies' own terms say

The companies that make chatbots address legal advice in their usage policies. These are contract terms set by each company, not laws, and each is dated below because they change.

• OpenAI, version effective October 29, 2025: users may not use OpenAI's services for "provision of tailored advice that requires a license, such as legal or medical advice, without appropriate involvement by a licensed professional."
• OpenAI, revised version labelled effective November 20, 2026 (scheduled, not yet in effect on November 6, 2026): the line becomes "building or offering AI tools that provide others with tailored advice that requires a license, without appropriate involvement by a licensed professional. You also may not state or imply that the AI itself holds professional credentials or authorization." The words "legal or medical advice" no longer appear in that line. The same page says, "Our rules are no substitute for legal requirements, professional duties, or ethical obligations that should influence how people use AI."
• Anthropic, version labelled effective November 12, 2026 (scheduled): for high-risk uses, including legal, "we require that a qualified person remain accountable for what is delivered or decided, and that the impacted individual knows AI was involved in the advice or decision."
• Google, Generative AI Prohibited Use Policy (last modified December 17, 2024): prohibits uses that make "automated decisions that have a material detrimental impact on individual rights without human supervision in high-risk domains," listing legal among the examples.

Did ChatGPT stop giving legal advice?

Not in the way the headlines suggested. The October 29, 2025 OpenAI wording restricts "tailored advice that requires a license" given "without appropriate involvement by a licensed professional." It does not prohibit discussing legal topics or explaining how the law works. The revised version scheduled for November 20, 2026 narrows that line: it bars "building or offering AI tools that provide others with tailored advice that requires a license" without a licensed professional's involvement, and it no longer names legal or medical advice. Both versions also separately bar automating high-stakes decisions ("actions" in the revised wording) in sensitive areas, including legal, without human review.

Are your AI chats private or privileged?

Do not assume they are. In United States v. Heppner (S.D.N.Y. No. 25 Cr. 503), Judge Jed S. Rakoff granted the government's motion from the bench on February 10, 2026 and explained why in a memorandum filed February 17, 2026. About 31 documents recording the defendant's own exchanges with Anthropic's Claude, created after he knew he was a target of a grand jury investigation, were not protected by attorney-client privilege or the work-product doctrine.

The court gave three main reasons. First, no attorney-client relationship existed: "No such relationship exists, or could exist, between an AI user and a platform such as Claude." Second, the exchanges were not confidential, because the platform's privacy policy allowed it to collect inputs and outputs, use them for training, and disclose them to third parties, including government regulatory authorities. Third, the documents were not work product because they "were not prepared at the behest of counsel."

The ruling has limits. It came from one federal trial court, in a criminal case, about a defendant acting on his own with a publicly available AI tool. The court did not decide how privilege applies when a lawyer directs a client to use an AI tool. The court also noted that it was unaware of any earlier case that had presented the issue.

Privacy policies point the same way. OpenAI's privacy policy (updated July 30, 2026) says it may share personal data, "including information about your interaction with our Services, with government authorities, industry peers, or other third parties in compliance with the law." The listed grounds include complying with a legal obligation (or a good faith belief that sharing is needed to comply with one), protecting its rights or property, a violation of its terms, policies or the law as it determines in its sole discretion, detecting or preventing fraud or other illegal activity, protecting safety and security, and protecting against legal liability.

The N.D. Cal. federal court's guidance for self-represented filers lists, as a risk of AI use, that it may "Be discoverable by a party in your case."

Lawyers face a separate duty. ABA Formal Opinion 512 (July 29, 2024) says a client's informed consent is required before a lawyer inputs information relating to the representation into a self-learning generative AI tool; that rule governs lawyers, not consumers using a chatbot on their own.

Can you trust what AI tells you about the law?

Not without checking. Courts have sanctioned lawyers and self-represented people who filed AI-generated briefs with citations to cases that do not exist or do not say what the brief claims. Our AI in Court guide covers court rules on AI, the leading sanctions cases and judges' own errors.

Two federal cases show what can happen to a self-represented litigant:

The Maine court noted that pro se litigants "must be held generally to the same standards as an attorney." Representing yourself does not lower the bar for accurate citations.

Using AI safely if you represent yourself

Courts that have addressed AI use generally allow it but put the responsibility on the filer. The Illinois Supreme Court's AI policy (effective January 1, 2025) says AI use by litigants and others "is authorized provided it complies with legal and ethical standards," and that "self-represented litigants are accountable for their final work product." New York court rules effective June 1, 2026 say AI use by attorneys and parties in preparing court papers "should not be prohibited."

The N.D. Cal. federal court's page for pro se filers says: "Using AI in arguing your case is permitted, though there are guidelines and rules that you must follow. Most importantly, regardless of which tools you use, you are responsible for everything you file with the Court." The California Fourth Appellate District's self-help page (October 31, 2025) puts it more bluntly: "AI is not a lawyer," and "You must make sure all court papers are correct, even if AI helped."

Practical steps that follow from those court pages and the cases above:

• Check every citation yourself. Open each case, statute and rule in an official source, such as the court's or legislature's own website, and confirm it exists and says what your paper claims.
• Read your court's rules and your judge's page. Some federal district courts have local rules or general orders on AI, and many individual judges have standing orders. State rules differ too; our AI in Court guide tracks them.
• Keep confidential details out of the chat. Do not paste information you would not want an opposing party or the government to see. After Heppner, assume a chat with a consumer AI tool is not privileged.
• Treat the output as a draft, not an answer. Deadlines, filing formats and local procedures vary by court, and a chatbot can state them wrongly with confidence.

When to get a human instead

Some problems carry stakes or deadlines where a mistake is hard to undo. In these situations, talk to a licensed attorney or a legal aid program before relying on anything an AI tool says:

• criminal charges, or any contact from police or prosecutors about an investigation;
• an eviction notice or a foreclosure;
• child custody, divorce, or a protective order;
• a serious injury claim, where filing deadlines apply;
• immigration matters;
• any court deadline that is close, or any paper you have been served.

A lawyer can review the specifics of your situation. Nothing on this page predicts how a court would decide your case.

What to Watch in 2027

These are the dated events found as of November 6, 2026. Some fall in late 2026 but shape what applies going into 2027:

• November 12, 2026: Anthropic's revised usage policy, labelled effective that date, requiring a qualified person to remain accountable for high-risk legal uses.
• November 20, 2026: OpenAI's revised usage policies, labelled effective that date, replacing the October 29, 2025 wording on tailored licensed advice.
• December 2, 2026: status hearing in Nippon Life v. OpenAI (N.D. Ill.). No date has been set for a ruling on OpenAI's motion to dismiss, which remained under advisement as of the docket read through October 8, 2026.
• New York S7263A: a 2025-2026 session bill committed to the Senate Rules Committee on June 5, 2026. That was its status on the Senate's bill page as of November 6, 2026.

We found no rule, statute or court decision on AI and the unauthorized practice of law scheduled to take effect in 2027, searching the court dockets, legislature pages and company policies cited on this page as of November 6, 2026.

State AI statutes are covered in our AI laws guide. Lawsuits claiming that chatbots harmed users are covered in our AI chatbot lawsuit guide.

Where to get real legal help

If you need help with a legal problem, these routes connect you with people, not software:

• Legal aid (civil cases): the Legal Services Corporation's Get Legal Help tool lets you "enter an address or city below to find an LSC-funded legal aid organization near you." LSC funds 129 independent nonprofit legal aid organizations. Federal rules generally cap eligibility at 125% of the federal poverty guidelines (45 CFR 1611.3(c)), and programs may set asset limits and adopt exceptions.
• Your state bar's lawyer referral service: many state and local bar associations run lawyer referral services that connect callers with licensed lawyers. Check your state bar's website.
• ABA Find Legal Help: the American Bar Association's directory points to free legal help, lawyer referral services and lawyer licensing lookups. The ABA itself says it "is not able to refer you to an attorney."
• Court self-help centers: many state courts run self-help centers and websites for people without lawyers. California's is the California Courts Self-Help Guide, which covers eviction, divorce, custody, restraining orders, small claims and more.
• 211: for non-legal essentials such as housing, food and utilities, call 211 or visit 211.org to reach local community services.

This page is general legal information about U.S. law on AI and legal advice, current as of November 6, 2026. It is not legal advice. For advice about your situation, talk to a lawyer licensed in your state.

Frequently Asked Questions

Can ChatGPT give legal advice?

ChatGPT can explain legal concepts, but it is not a lawyer and cannot represent you. OpenAI's usage policies in effect since October 29, 2025 bar using its services for tailored advice that requires a license, such as legal advice, without appropriate involvement by a licensed professional; a revised version labelled effective November 20, 2026 narrows that line to people who build or offer AI tools that give others tailored advice requiring a license. You remain responsible for checking anything it says.

Is it illegal for AI to give legal advice?

As of November 6, 2026, we found no court ruling, statute or bar opinion holding that an AI tool itself practices law. New York's unauthorized-practice statutes, for example, are written for a 'natural person', and a pending federal case, Nippon Life v. OpenAI in the Northern District of Illinois, alleges ChatGPT practiced law without a license; OpenAI's motion to dismiss had not been decided as of the docket read through October 8, 2026.

What counts as legal advice?

Each state defines the practice of law through its own statutes and courts. As a working line, legal information explains what the law says in general, while legal advice applies the law to a specific person's facts and recommends what to do, such as which claims to file or how to complete papers to sign or file.

Did ChatGPT stop giving legal advice?

Not in the sense of refusing legal topics. OpenAI's October 29, 2025 usage policy bars tailored advice that requires a license, such as legal advice, without a licensed professional's involvement. A revised version labelled effective November 20, 2026 narrows that line to building or offering AI tools that give others tailored advice requiring a license, and no longer names legal advice in it.

Are my ChatGPT chats protected by attorney-client privilege?

Do not assume so. In United States v. Heppner, the federal court in the Southern District of New York ruled in February 2026 that a criminal defendant's own exchanges with an AI platform, Anthropic's Claude, were not privileged because no attorney-client relationship can exist with an AI platform and the chats were not confidential. The ruling did not decide use directed by a lawyer.

Can I use ChatGPT to write court papers?

Courts that have addressed it, including the Illinois and New York state courts and the federal court in the Northern District of California, generally allow it, but you are responsible for everything you file. Federal courts have sanctioned self-represented litigants over false citations, for example a $2,900 Rule 11 sanction for 29 false citations against a pro se defendant in a Michigan federal case in December 2025. Check every citation in an official source and read your court's rules and your judge's standing orders.

Where can I get free legal help?

For civil problems, the Legal Services Corporation's Get Legal Help tool finds LSC-funded legal aid near you; eligibility is generally capped at 125% of the federal poverty guidelines, with program exceptions. State bar lawyer referral services and state court self-help centers are other options, and 211 connects you with non-legal community services.