Your Chatbot is Not a Lawyer!

Every day, millions of us quietly interact with artificial intelligence whether we know it or not. Often, it’s not even obvious that we’re relying entirely on AI to manage whatever we’re doing online. In many cases, it’s intentional, such as letting ChatGPT polish our resumes, having Gemini generate a quick summary of news events, or asking Claude to curate our social feeds.
We often get calls after the potential client looked for advice from an AI chatbot on how to respond to the investigation or case filed against them or the potential legality of their conduct. Unfortunately, those AI chats, including any details or information provided to the chatbot, may have to be turned over in responding to the investigation or litigation they are facing. For many years, the government has made requests for Google (Yahoo, Bing, etc) search history. Now, we are seeing SEC subpoenas requiring production of AI chats.

While government investigators cannot compel an attorney to disclose communications other than where the attorney is assisting in illegal conduct, those investigators can compel communications with AI chatbots concerning the law and potential liability. Yet, the ease of “conversing” with AI chatbots – rather than calling an attorney – means many people are going to create evidence (chats) that the government can and will use against them.

Your AI Chat History is not Confidential
You should be aware that records of your interactions with AI chatbot services — including, but not limited to, platforms such as ChatGPT (OpenAI), Claude (Anthropic), Gemini (Google), Copilot (Microsoft), and any other generative AI tools —are definitely not protected by any privilege like the attorney-client privilege or the physician-patient privilege. The Securities and Exchange Commission, FINRA, or a host of other federal and state regulatory bodies can and do request AI chatbot conversations. This includes the questions you ask, responses generated by the AI, uploaded documents, conversation histories, and any associated metadata, including timestamps, IP addresses, and account information.

Government investigators may obtain these records through several mechanisms, including grand jury subpoenas, administrative subpoenas, search warrants, court orders, or civil investigative demands. If you have any kind of license, you may be obliged to turn over the records to the government or regulatory agency at any time they ask for the information. In fact, the request might not even be directed to you but to the AI service providers themselves. Many AI platforms retain user data pursuant to their terms of service and privacy policies, and these companies are generally obligated to comply with valid legal requests.

Why This Matters

AI chat histories can be remarkably revealing. Users frequently interact with AI tools in a candid, unguarded manner — treating them as confidential sounding boards. As a result, your AI chat history may contain statements that can be used against you such as:
• Evidence of knowledge, intent, or motive relevant to the investigation;
• Drafts of communications, documents, financial analyses, spreadsheets, or data that you may have shared with the AI for analysis;
• Requests for legal or strategic advice that reveal your state of mind; and
• Searches or inquiries that could be characterized as evidence of consciousness of guilt or efforts to conceal information.

Unlike a conversation with an attorney, communications with an AI chatbot are not protected by the attorney-client privilege. There is no recognized legal privilege shielding AI chat interactions from disclosure and any information you share with an AI platform should be treated like search engine histories.

A few courts have held that a party’s use of third-party AI tools can be protected not by attorney-client privilege, but by the “work product privilege” which can apply to the parties themselves. But many courts have refused to recognize any privilege to protect AI search communications. These few cases do not deter the SEC and other agencies from requesting AI chats. If you refuse to produce the AI chat history, the SEC can and likely would file a public subpoena enforcement action in federal court seeking a court order requiring production. That would make the nonpublic investigation a very public one.

Immediate Preservation Obligation

Importantly, once you have notice of the investigation or litigation, you must take steps to preserve all AI chat records in your possession, custody, or control, just like you need to do with all other documents, data and records. The intentional destruction of potentially relevant evidence — sometimes referred to as spoliation — can result in severe legal consequences, including adverse inference instructions, monetary sanctions, contempt findings, and, in the context of a government investigation, separate criminal charges for obstruction of justice or destruction of evidence.

Recommended Immediate Steps

We strongly recommend that you take the following actions if you receive an inquiry from a regulator or believe one is coming:

1. Do not use AI chatbots to discuss, analyze, or seek advice regarding the investigation or any related matters. Any such communications would almost certainly be discoverable and could be used against you.
2. Contact our office immediately to schedule a confidential consultation. Unlike AI chatbots, communications with an attorney – even those you have with an attorney you are consulting before retaining them – are protected by the attorney-client privilege.