Law Firms Warn: Sharing Attorney Communications With AI Chatbots May Waive Privilege

By Hector Herrera | April 26, 2026 | Legal

Major U.S. law firms have issued urgent client advisories with a clear warning: sharing legal communications with AI chatbots could waive attorney-client privilege — the foundational protection that keeps confidential communications between lawyers and clients out of court proceedings. The warning follows a ruling in which a litigant's ChatGPT conversations were deemed potentially discoverable by opposing counsel, according to Claims Journal coverage published April 16.

The attorney-client privilege issue is not theoretical. It is being litigated right now, with conflicting rulings already on the books.

What Happened

A litigant involved in active legal proceedings used ChatGPT to assist with case preparation — inputting information about the dispute, asking questions about legal strategy, and reviewing draft arguments. When opposing counsel requested discovery of those AI conversations, a court ruled the logs were potentially discoverable. The privilege protection that would have applied to communications with a human attorney did not automatically extend to the chatbot interactions.

In a separate, conflicting ruling, a Michigan federal court treated a pro se litigant's (a person representing themselves without an attorney) AI chat logs as personal work product — affording them some protection from discovery. But the protection granted to work product is narrower and more conditional than attorney-client privilege, and it applies differently to represented parties.

Two courts. Two different outcomes on closely related fact patterns. That is the current state of AI-in-litigation legal doctrine.

Why Privilege May Not Attach

Attorney-client privilege protects confidential communications made for the purpose of seeking legal advice from a licensed attorney. The doctrine rests on several requirements:

1. The communication is confidential — not shared with third parties outside the privilege
2. The communication is with a licensed attorney — or the attorney's agents and employees acting in that capacity
3. The purpose is seeking legal advice — not business or strategic advice generally

AI chatbots fail on multiple dimensions. They are not licensed attorneys. Communications shared with them are, depending on the platform's terms of service, potentially stored, used to train future models, or accessible to the platform provider — none of which qualifies as confidential in the legal sense.

When a client shares privileged communications with an AI chatbot — forwarding an email from their attorney, pasting attorney memos, describing confidential legal strategy — they may be treating a third party as inside the privilege relationship when it legally is not.

What Law Firms Are Telling Clients

The advisories circulating from major firms center on several practical warnings:

• Do not input privileged communications into AI chatbots. This includes attorney emails, memos, legal strategies, and case assessments
• Do not describe confidential legal matters in AI prompts even in general terms, if the description contains information that could identify the matter or the parties
• Check platform terms of service. Some enterprise AI tools offer contractual confidentiality that may provide some protection; consumer products like the free tiers of ChatGPT, Claude, and Gemini generally do not
• Assume AI conversations are discoverable until courts clarify the doctrine or legislation establishes protections

For in-house legal departments, the advice extends to AI tools embedded in corporate systems. If legal counsel is using an AI tool to assist with privileged work product and that tool shares data with a vendor's servers, the privilege question applies.

The Broader Pattern: AI in Legal Practice Is Outpacing Legal Doctrine

The privilege question is one part of a broader pattern in which AI use in legal practice is advancing faster than the legal doctrine governing it.

Courts have already sanctioned attorneys for submitting AI-generated briefs containing fabricated case citations — the "hallucination" problem. AI-generated discovery analysis, AI-assisted contract review, and AI-drafted filings are all in active use at law firms of every size. The professional responsibility rules governing attorney competence, candor to the court, and confidentiality were written in a pre-AI world.

The American Bar Association and multiple state bar associations have issued ethics opinions acknowledging that AI use in legal practice must be governed by existing professional responsibility rules — but "existing rules apply" is not the same as clear guidance on how they apply in novel AI-specific scenarios.

What Clients and Attorneys Should Do Now

For clients in active litigation or with ongoing legal matters:

• Treat AI chatbots as you would treat any third-party service provider outside the privilege: don't share confidential attorney communications with them
• If you use AI to help you understand legal concepts or draft communications that are not yet attorney communications, that is a lower-risk use
• Ask your attorney explicitly about your firm's AI tool policies and what protections are in place

For attorneys:

• Assess whether your firm's AI tool contracts include confidentiality protections sufficient to preserve privilege
• Create written client disclosures about AI tool use in their matters
• Document the human attorney review and judgment applied to any AI-assisted work product

What to Watch

Watch for bar association guidance to become more specific about AI and privilege preservation in the second half of 2026. A model rule or formal ethics opinion clarifying the doctrine would give courts, attorneys, and clients a clearer framework.

Congressional proposals for AI transparency and user data protections could also create new categories of confidentiality protection for AI interactions — though that legislative path is uncertain.

Source: Claims Journal