Can AI Give Legal Advice?

No. AI does not give legal advice. It gives legal information. That difference decides every other answer on this page. Legal advice means a licensed lawyer applies the law to your facts inside a relationship where they owe you a duty. AI is not a lawyer, forms no attorney-client relationship, and owes you nothing. So treat any AI legal output as information to verify with a lawyer, never as advice you can act on. This is general information, not legal advice. Checked July 2026.

TL;DR

• AI gives legal information, not legal advice. Information is a general explanation of the law. Advice is the law applied to your facts by a licensed lawyer who is responsible to you.
• Only a lawyer in a relationship gives advice. Legal advice is part of the practice of law. Doing it without a license is unauthorized practice of law (UPL).
• No relationship, no duty, no privilege. A chatbot is software run by a vendor. There is no attorney-client relationship, no malpractice duty, and no privilege over what you type.
• The danger is that AI sounds like advice. A confident answer in clean prose reads exactly like counsel. It is not. That gap is where people get hurt.
• AI is safe for some tasks, not others. It can explain concepts, draft from your input, and summarize text you give it. It cannot resolve your specific situation or be relied on alone.
• For lawyers, AI is a tool, not a delegate. You stay responsible for the output. For everyone else, AI is a starting point, not a substitute for a lawyer.

Legal information vs legal advice: the line that decides everything

Start with the distinction, because every other answer flows from it.

Legal information is a general explanation of how the law works. "A non-compete must usually be reasonable in scope and duration" is information. It is true in the abstract, it is not about you, and anyone can say it. General legal information is broadly available and lawful to share.

Legal advice is the law applied to your specific facts, with a recommendation about what you should do. "Your non-compete is unenforceable in your state, so you can take the job" is advice. It depends on your facts, your jurisdiction, and a judgment call about your risk.

AI is very good at the first thing and structurally incapable of the second. It can produce a tidy summary of non-compete doctrine. It cannot owe you a duty to get your specific answer right.

Can AI give legal advice under US law? The UPL problem

Legal advice is more than a label. It is part of the practice of law, and the practice of law is regulated.

Every US state restricts who may practice law to licensed attorneys. Giving legal advice, advising on a person's rights, and drafting documents that affect those rights are classic examples of practicing law. A non-lawyer who does these things commits unauthorized practice of law (UPL). The rules exist to protect the public from unqualified advice and to keep someone accountable when it goes wrong.

This is not a new fight. For years, online legal-form providers like LegalZoom faced UPL challenges arguing that automated document assembly crossed into practicing law. Those cases mostly settled or resolved by drawing a line between selling self-help forms and giving tailored advice. AI raises the same question with more force. A chatbot does not just fill in a template. It answers your specific question in natural language, which looks far more like advice than a fill-in-the-blank form ever did.

So the question "can AI give legal advice" is partly a question about UPL. If a tool actually delivered tailored legal advice to the public, that would look like the practice of law without a license. The current answer from the tools themselves is to stay on the information side of the line and disclaim the rest.

OpenAI made this explicit. In a 2025 update to its usage policies, it barred using the service for "tailored advice that requires a license, such as legal or medical advice, without appropriate involvement by a licensed professional" (OpenAI Usage Policies). The boundary the policy draws is the same one the law draws: tailored advice needs a licensed human in the loop.

Why AI output is not advice: no relationship, no duty, no privilege

Even when AI gives you a confident, specific-sounding answer, three things are missing that make real advice what it is.

No attorney-client relationship. Advice happens inside a relationship you form with a lawyer. A chatbot cannot form that relationship. It is a service run by a company, not a member of any bar. Nothing you type creates a client.

No duty to you. A lawyer who advises you owes duties of competence and loyalty, and can be liable for malpractice if they get it badly wrong. A model owes you nothing. If its answer is wrong, there is no professional on the hook. The vendor's terms push the risk onto you.

No privilege. Attorney-client privilege protects confidential talk between a client and a lawyer seeking advice. A chatbot is neither, so privilege never attaches, and pasting sensitive facts can even waive protection you held. We cover this in does ChatGPT have attorney-client privilege and is ChatGPT confidential for legal work.

This is also why the vendors disclaim. Their terms warn you not to rely on output as a substitute for professional advice. Commentators read the legal-advice carve-out as a liability shield as much as a public-safety move, shifting risk to the user when things go wrong (Bloomberg Law).

The dangerous gap: AI sounds exactly like advice

Here is the part most people miss. The risk is not that AI refuses to help. The risk is that it sounds like a lawyer when it is not one.

Ask a model about your eviction, your contract, or your custody fight, and it will answer in calm, confident, well-organized prose. It reads like counsel. It feels authoritative. None of that polish means the answer is right for your facts, your state, or this year.

Independent testing backs this up. A Stanford RegLab study (Magesh et al., 2025) found that even purpose-built legal research tools from LexisNexis and Thomson Reuters hallucinated between 17% and 33% of the time, and general chatbots did worse (Stanford RegLab). A wrong answer arrives in the same confident tone as a right one, and the model rarely flags the doubt. You cannot tell which is which from the writing alone.

The mistake we see is rarely someone asking a model what a statute means in the abstract. It is a user pasting their own lease, offer letter, or operating agreement, getting a confident answer, and treating the risk as handled. The general explanation was probably fine. The specific reliance is where it goes wrong.

What AI can safely do, and what it cannot

The honest split is by task. AI earns its keep where you supply the facts and the law, and you do the relying.

SAFE (information / drafting from your input)      NOT SAFE (advice / reliance)
-------------------------------------------        ----------------------------------
Explain a legal concept in plain English           Tell you what to do in your case
Summarize a document you paste in                   Decide if your contract is enforceable
Draft a first version from your facts               Confirm a deadline you will rely on
Build a checklist of questions to ask               Replace asking a licensed lawyer
List general pros and cons of an option            Give a final answer you act on alone

What AI can do well:

• Explain concepts. What is an indemnity clause, how does at-will employment work, what is the difference between a trademark and a copyright. General information, safely.
• Draft from your input. Hand it your facts and ask for a first-draft demand letter, policy, or summary. You then edit and verify.
• Summarize text you provide. Paste a long agreement or thread and get the key points. It is working from your document, not its memory.

What AI cannot safely do:

• Advise on your specific situation. It cannot weigh your facts and your jurisdiction and tell you what to do, because it owes you no duty and may be wrong.
• Be relied on without a lawyer. Its output is a draft or an explainer, not a verified, accountable answer.
• Find or confirm the law for you. General chatbots fabricate citations and miss recent changes, so any legal proposition needs checking against the source.

Consumers vs lawyers: same tool, different posture

The "can AI give legal advice" question lands differently depending on who is asking.

For a member of the public, AI is a way to get oriented, not a lawyer. It can help you understand a concept or organize your questions before a consultation. It cannot replace a licensed attorney who owes you a duty. The cheapest mistake is treating a confident chatbot answer as legal advice you can act on. The fix is to use it to prepare, then talk to a real lawyer.

For a lawyer, AI is a tool in the workflow, and the responsibility never leaves you. ABA Formal Opinion 512, issued July 2024, applies the Model Rules to generative AI. Its core point is that the lawyer stays responsible for competent service, so uncritical reliance on AI output without an appropriate degree of independent verification can breach the duty of competence (ABA Formal Opinion 512). The AI does not advise the client. The lawyer does, using AI as one input. We break the opinion down in our ABA Formal Opinion 512 guide and cover safe day-to-day use in ChatGPT for lawyers.

The throughline is the same for both. AI produces information. A licensed lawyer turns information into advice and stands behind it.

The verdict

Can AI give legal advice? No, and you should not want it to.

What AI can do is give you legal information fast: explain a concept, draft from your facts, summarize your documents. That is genuinely useful, and for a lawyer it is a real productivity tool. What it cannot do is form a relationship, owe you a duty, or take responsibility for an answer about your specific situation. Those are the things that make advice advice.

So the practical rule is simple. Use AI to understand and to draft. Use a lawyer to decide and to rely on. When the output sounds like advice, remember it is not, and verify with someone who is accountable to you.

A grounded legal tool narrows the gap on the information side. Vaquill AI is a legal AI suite for in-house counsel that retrieves over real US statutes and opinions, so each cite links back to a source you can open and check, not a guess you have to trust. One scope note: our public API is statutes-only (US Code, CFR, and 50-state codes), and case-law research lives inside the product. It still does not replace a lawyer's judgment. It just makes the verification a lawyer owes far cheaper. For a fuller roundup, see the best legal AI tools for in-house counsel.

FAQ

Can AI give legal advice? No. AI gives legal information, which is a general explanation of the law. Legal advice is the law applied to your specific facts by a licensed lawyer inside a relationship where they owe you a duty. AI is not a lawyer and cannot form that relationship, so its output is information to verify, not advice to act on.

What is the difference between legal information and legal advice? Legal information is general and not about you, like "non-competes must usually be reasonable." Legal advice is specific and about you, like "your non-compete is unenforceable, so you can take the job." Information is broadly lawful to share. Advice is part of the practice of law and reserved for licensed attorneys.

Is it illegal for AI to give legal advice? The practice of law, including giving tailored legal advice, is reserved for licensed attorneys, and a non-lawyer doing it can commit unauthorized practice of law (UPL). To stay clear of that line, AI tools generally provide information and disclaim advice. OpenAI's usage policy now bars tailored advice that requires a license without a licensed professional involved.

Does talking to an AI create an attorney-client relationship? No. A chatbot is software run by a vendor, not a member of any bar, so no attorney-client relationship forms. That means no duty of competence or loyalty to you, no malpractice accountability, and no attorney-client privilege over what you type.

Can I rely on AI legal answers if they sound confident? No, and the confidence is the trap. AI answers legal questions in calm, authoritative prose whether the answer is right or wrong, and general models get a large share of legal questions wrong. Treat the output as information to confirm with a lawyer, never as a final answer you act on alone.

Can lawyers use AI to give legal advice to clients? Lawyers can use AI as a tool, but the lawyer gives the advice and stays responsible for it. ABA Formal Opinion 512 says uncritical reliance on AI output without appropriate independent verification can breach the duty of competence. The AI is one input, the licensed lawyer is the one who advises and stands behind it.

What can AI safely do with legal questions? It can explain legal concepts in plain English, draft a first version from facts you supply, and summarize documents you paste in. These are information and drafting tasks where you provide the substance and do the verifying. It cannot resolve your specific situation or be relied on without a lawyer.

Should I use ChatGPT or a lawyer for my legal problem? Use AI to understand the issue and prepare your questions, then bring in a lawyer for the decision you will rely on. AI can orient you cheaply and fast. Only a licensed attorney can apply the law to your facts, owe you a duty, and be accountable for the answer.

Last updated: July 2026.

Arshita Anand

Co-Founder & CEO · Attorney

Arshita leads product and strategy at Vaquill, building the legal AI suite that solo, small-firm, and in-house US lawyers use to run a matter end to end.