What you type into an AI chatbot isn't protected - not even if you're using it to prepare for a legal case. That's the takeaway from a ruling in US v. Heppner by Judge Jed Rakoff of the Southern District of New York, issued in April 2026.
The ruling established that conversations with AI chatbots don't qualify for attorney-client privilege - the legal protection that prevents lawyers from being forced to reveal what their clients told them. The reasoning is direct: privilege covers communications between a person and their attorney. An AI chatbot is a third-party software tool, not an attorney. Even if a lawyer recommended using one to gather information or draft documents, the chat logs don't inherit any protection.
Following the ruling, US lawyers began warning clients to treat any AI chatbot conversation as potentially discoverable - meaning prosecutors, opposing counsel, or regulators could subpoena and read the full transcript. That applies to tools like ChatGPT and Claude regardless of what you were using them for: thinking through a legal situation, drafting documents, or asking questions you'd normally only ask a lawyer.
What "Discoverable" Actually Means
In civil and criminal cases, discovery is the process where each side can demand documents, messages, and records from the other. Emails, Slack chats, and text messages are routinely swept up. AI chat logs now sit in that same category.
The specific risk: if you typed details about an ongoing dispute, a business deal gone wrong, or anything else relevant to litigation into a chatbot, that transcript could end up as evidence. Chatbots also don't delete conversations by default - they store them on company servers, which can receive independent subpoenas regardless of anything you do on your end.
Three Things to Do Now
The practical guidance lawyers are giving after this ruling:
- Don't use AI chatbots to discuss active legal matters, pending litigation, or anything you'd only tell your actual attorney
- If you want AI assistance on legal documents, do it through systems your lawyer controls or through platforms with explicit confidentiality agreements in place
- Audit your chat history - both ChatGPT and Claude let you disable history logging or delete past sessions
This ruling comes from one federal court in New York, so it isn't binding on courts across the country yet. But the underlying logic is solid and the decision will be cited going forward. The more courts look at this question, the less likely it gets that chatbot conversations will ever be treated as privileged communications.