Federal Court Rules Public AI Chatbots Cannot Claim Attorney-Client Privilege
A New York federal court has issued its first ruling on whether conversations with public AI chatbots qualify for attorney-client privilege. The answer is no.
The decision carries immediate implications for in-house counsel and legal teams. If your company pastes legal advice, investigation materials, or other sensitive information into a public AI tool, you are disclosing it to a third party-and losing protection in the process.
What the Court Found
A federal investigation target had used a public AI platform to create strategy reports about his case. When agents seized his devices, he claimed the AI conversations were protected by attorney-client privilege because he later shared the outputs with his lawyers.
The court rejected the claim on three grounds:
- The AI is not your lawyer. Privilege protects confidential communications with an attorney. A chatbot is not an attorney, and no attorney-client relationship existed.
- There was no reasonable expectation of confidentiality. The AI provider's terms of service allowed the company to collect user inputs, train its models on them, and disclose them to third parties-including regulators.
- The chats were not about obtaining legal advice from counsel. The defendant initiated the conversations on his own. The AI tool disclaimed providing legal advice. This falls outside what privilege is designed to protect.
The court also rejected work product protection. That doctrine shields a lawyer's thinking and strategy. These documents were created by the client alone, using a public tool, not by or at the direction of counsel.
The court made one more critical point: information that starts as privileged loses that protection the moment it enters a public chatbot. Sharing privileged material with a third party constitutes waiver.
The Risk Extends Beyond Criminal Cases
The ruling applied narrow facts, but its logic extends to routine legal scenarios. Litigation discovery, regulatory investigations, enforcement actions, and internal investigations all depend on keeping certain information confidential.
Employees are already having unprotected conversations with AI tools daily. They ask about compliance obligations, draft internal policies, assess whether security incidents constitute data breaches. None of these conversations involve an attorney. None carry a reasonable expectation of confidentiality. Under the court's framework, all of them could become exhibits in litigation or regulatory proceedings.
Expect opposing parties and regulators to ask pointed questions. "Did you use any AI tools to determine your data privacy compliance obligations?" or "Provide all communications with AI-based tools related to investigation or remediation of the data breach" are no longer hypothetical discovery requests.
Paid Tools and Premium Subscriptions Don't Fix the Problem
The court focused on the specific AI provider's privacy policy. But its reasoning applies broadly. Any AI platform-free, paid, or commercially licensed-presents the same risk if its terms reserve the right to review, train on, or disclose user data.
A corporate license does not automatically solve the confidentiality problem. The door remains unlocked if the vendor's terms allow data access or model training.
Three Steps to Protect Privileged Information
Set clear rules and explain the stakes. Create a straightforward policy: employees may not input legal advice, attorney communications, investigation materials, draft policies, audit findings, or trade secrets into any unapproved AI tool. A concrete rule is far more effective than a vague instruction to "use caution."
Deploy enterprise AI with contractual safeguards. Use an enterprise solution that prevents model training on your data, blocks vendor access and disclosure, restricts use of your information in outputs for other customers, and keeps interactions in your controlled environment. Review and negotiate vendor agreements for confidentiality, data segregation, and audit rights. The court hinted that attorney direction might have changed the outcome. Require that any AI-assisted work involving legal content happen only under counsel's direction, within a documented workflow built for privileged communications.
Train employees to pause before pasting. Every prompt and uploaded document is a potential disclosure to a third party, regulator, or investigator. Before anyone uses a chatbot, they should ask whether the content is privileged, confidential, or sensitive. If yes-or even maybe-they should consult legal first. Restrict use of unapproved systems on company or personal devices. Run periodic scenario-based training so the rules stay reinforced.
The Bottom Line
This ruling did not create new law. It applied longstanding privilege principles to a new technology. But that is precisely what makes it significant. The court confirmed that public AI tools are third parties, that sharing information with them waives privilege as easily as forwarding a confidential email to a stranger, and that later attorney involvement cannot undo the disclosure.
Companies that act now can continue using AI productively. Those that do not are gambling that employees will never leave the door open-and this ruling shows exactly what happens when they do.
Learn more about AI for Legal and how to implement governance frameworks in your organization. For paralegals and legal support staff, the AI Learning Path for Paralegals covers document handling, privilege protection, and AI governance in legal workflows.
Your membership also unlocks:
