Companies like OpenAI, Anthropic and Google have promised next generation tools that can solve just about any problem. Naturally, one of the major questions people are asking is: should I use AI for legal advice?<\/p>\n
Absolutely not. Here’s why.<\/strong><\/p>\n The American Bar Association (ABA) has warned lawyers to treat generative AI as a tool that still requires competence and careful supervision. Federal courts in Texas<\/a> and elsewhere have also warned that AI cannot<\/strong> replace legal judgment. And multiple real cases<\/a> have already ended with sanctions, fines, fee awards, removal from cases, and disciplinary referrals after people relied on AI-generated legal content that turned out to be false.<\/p>\n States like Texas, California and Nevada have also recently passed laws<\/a> related to the use of AI chatbots as replacements to lawyers, doctors and therapists, warning that any harm caused as a result of those tools falls squarely on the heads of the organization using them.<\/p>\n ChatGPT can be useful for general information, or for finding a lawyer<\/a>, but it should not be used as your source of legal advice or for handling your case.<\/strong><\/p>\n <\/a><\/p>\n Courts, ethics authorities, and even recent sanctions orders all point in the same direction: AI has been known to invent facts, create new cases, and misstate the law. Multiple billion dollar companies like Deloitte<\/a>, as well as some lawyers, have been fined for the presence of fake law and case citations.<\/p>\n The truth is that you don’t know what it is you don’t know<\/strong>. All it takes is one bad citation, one phony law, or one badly contructed or misunderstood argument for the defense to be dismissed from the courthouse. Using AI as your lawyer means taking on a host of responsibilities that most people do not consider, including:<\/p>\n A chatbot does not inspect a crash scene, subpoena records, challenge an adjuster, or stand up in court and take responsibility for what it told you. “Good enough\u201d legal information is not good enough when your life and livelihood are on the line. Real legal advice should come from a real lawyer.<\/p>\n <\/a><\/p>\n The biggest problem with AI is not that it is always wrong. It is that it can be almost right <\/em>in a way that sounds polished, confident, and complete.<\/p>\n A personal injury victim may ask a chatbot whether a trucking company can be liable, whether a recorded statement is required, or whether a settlement offer is \u201cgood.\u201d Those are not generic questions. They depend on facts the AI may not know, facts the user may not realize are important, or legal rules that may vary by jurisdiction.<\/p>\n This is not only a lawyer problem. In January 2026, the Seventh Circuit addressed a pro se litigant\u2019s filing that contained quotations not found in the cited cases. The court stopped short of sanctions, but said it was skeptical the brief had not been prepared with AI and stressed that accuracy and honesty matter for unrepresented parties too. The court wrote that pro se litigants (non-lawyers) \u201cshoulder responsibility\u201d and must take reasonable care to avoid factual and legal misrepresentations.<\/p>\n So even if someone is not hiring a lawyer and is trying to use AI on their own, the courts still expect real-world care and accuracy. It also means that any problems with filings, briefs or other generated documents become your responsibility – including any fees or fines levied by the court.<\/p>\n <\/a><\/p>\n One of the first major AI-sanctions cases came in June 2023 in Mata v. Avianca. In that case, lawyers opposing dismissal in an airline injury case submitted authorities that did not exist. The federal judge imposed a $5,000 sanction, ordered the lawyers to notify their own client, and even ordered them to send letters to judges falsely identified as authors of the fake opinions.<\/p>\n Just last February, a federal judge in Wyoming sanctioned three lawyers<\/strong> in a personal injury case against Walmart after fake AI-generated cases appeared in a filing. Reuters reported that the fines totaled $5,000, one lawyer was removed from the case, and the court emphasized that the duty to verify sources \u201cremains unchanged\u201d even as lawyers move into the AI era.<\/p>\n That case involved a real personal injury lawsuit over an allegedly defective hoverboard that allegedly caused a house fire.<\/p>\n On February 18, 2026, the U.S. Court of Appeals for the Fifth Circuit sanctioned an attorney after finding that she used AI to draft a substantial portion, if not all, of a reply brief and failed to verify its accuracy. The court imposed a $2,500 sanction. The order also says the attorney was not forthcoming in responding to the court\u2019s show-cause order.<\/p>\n The Sixth Circuit went even further in Whiting v. City of Athens<\/em>. In a published opinion issued March 13, 2026, the court said the lawyers\u2019 briefing repeatedly misrepresented the record, cited nonexistent cases, and cited cases for propositions they did not support. The court ordered the lawyers to reimburse the appellees\u2019 full reasonable appellate attorney\u2019s fees, pay double costs, and pay $15,000 each in punitive sanctions. It also directed the clerk to send the order to the chief judge for possible disciplinary proceedings.<\/p>\n <\/a><\/p>\n Multiple recent cases have resulted in dismissals due to the use of AI that led to hallucinations, inaccuracies and fabricated quotations. The following examples involve dismissals without prejudice,\u00a0<\/strong>meaning that the plaintiffs can bring the case again. All of these cases represent a growing frustration with the use of generative AI in court.<\/p>\n ByoPlanet line of cases (S.D. Fla., July 15, 2025) <\/strong>– The court found an attorney used generative AI to develop hallucinated cases and fabricated quotations. The attorney also continued doing so after being put on notice. The judge then dismissed all four cases without prejudice and without leave to amend. The court also ordered fee shifting and referred counsel to the Florida Bar.<\/p>\n Dubinin v. Papazian (S.D. Fla., Nov. 21, 2025) <\/strong>– The court found the plaintiff\u2019s filings contained AI-generated inaccuracies, including nonexistent authorities and false quotations. The court struck the case and dismissed it without prejudice. The court also ordered the plaintiffs to pay attorneys\u2019 fees for the defense.<\/p>\n <\/a><\/p>\n Your chats with AI tools are not private. Whatever you tell or share with ChatGPT, Gemini or any other chatbot is not\u00a0<\/strong>federally protected evidence. Just two months ago, on Feb. 10, 2026, the U.S. District Court for the Southern District of New York issued a ruling<\/a> that chats with generative AI are not protected by attorney-client privilege. This means that the opposing party can request any\u00a0<\/strong>evidence shared with AI. Lawyers work hard to only present evidence that helps your case to the opposing counsel or the court. However, when you use AI, the other side can request all\u00a0<\/em>of the information that you share with the chatbots.<\/p>\n <\/a><\/p>\n This article should not be taken to say that AI should never be used at all. Used carefully, AI can be helpful. It may help you understand your case, or legal terms that you might not understand, like negligence, comparative fault, liens, subpoena and so on. It can help you organize a timeline, make a list of questions for a lawyer, or summarize documents for your own review. All of this is extremely helpful when starting a lawsuit. Additionally, chatbots have grown to be very strong search engines, so asking AI about a particular lawyer can help make a decision on whether or not to contact their office.<\/p>\n However, all of these are different from telling you what you should<\/strong> do in your case, or from litigating it for you. The American Bar Association\u2019s Formal Opinion 512 makes clear that even lawyers using generative AI still have to satisfy basic legal duties. These include competence, confidentiality, and communication. In other words, the legal profession itself treats AI as a tool for experienced attorneys, not something that replaces judgment outright.<\/p>\n
\n\n
\nAI Is Not a Lawyer<\/h2>\n
\n
Why AI Is Risky for Legal Advice<\/h2>\n
Cases Involving AI Use Resulted in Sanctions<\/h2>\n
Mata v. Avianca (2023)<\/h3>\n
The Wyoming Walmart case<\/h3>\n
The Fifth Circuit\u2019s February 18, 2026 sanction order<\/h3>\n
The Sixth Circuit\u2019s March 13, 2026 sanctions<\/h3>\n
Use of AI Resulted in Case Dismissals<\/h2>\n
AI Does not Respect Privacy<\/strong><\/h2>\n
What AI Can Help With<\/h2>\n