Are AI Legal Chats Protected by Attorney-Client Privilege?

Do not assume that a conversation with an AI legal chatbot is protected by attorney-client privilege. A private account, a helpful answer and a legal topic do not establish that protection. The answer depends on the governing law, the purpose and circumstances of the communication, the lawyer-client relationship and how information is shared.

This U.S. overview helps you identify the questions to ask before entering sensitive material. It does not decide whether a particular conversation is privileged. If an existing dispute, investigation or disclosure is involved, ask your lawyer about the actual records and applicable law before taking further action.

Four protections that should not be collapsed into one

These concepts often appear together, but they answer different questions:

The ABA's Model Rule 1.6 commentary expressly distinguishes a lawyer's ethical confidentiality responsibilities from evidentiary attorney-client privilege and work product. The ethical duty can extend beyond information a client personally communicated. These are model principles; the rules adopted in the relevant jurisdiction govern.¹

This distinction matters when a service says information is “secure.” Security controls may be relevant to risk, but they do not themselves answer an evidentiary privilege question. Similarly, a lawyer's duty to safeguard information does not establish that every document in a lawyer's possession is privileged.

Independent chatbot use needs a separate assessment

A person asking a third-party chatbot for an explanation of a lease should not assume they are communicating with a lawyer. The Northern District of California's AI guidance warns that information entered into or received from third-party tools may be stored outside the user's control, lack confidentiality or privilege, and be discoverable.²

Treat this as a reason to pause before disclosure, not as a claim that every AI interaction in every context has an identical legal result. Product terms, access arrangements and legal context can differ. A label such as “legal assistant,” “private workspace” or “enterprise” does not resolve those differences on its own.

Start with a general or fictional question where possible. Asking what a term commonly means is different from uploading a complete attorney email, settlement strategy or medical record. If the real document is necessary for useful advice, the next step may be a lawyer-approved channel rather than more detailed prompting.

A lawyer's approved tool is not an automatic guarantee

A law firm may evaluate an AI service for a defined task. That review can include data use, retention, access, security and the need for client communication or informed consent. ABA Formal Opinion 512 discusses these professional duties and the fact-specific risks of disclosing representation-related information to a tool.³

Nevertheless, “my lawyer uses it” is not enough to answer whether a specific communication is privileged. Ask what the lawyer has approved, for which purpose and account, and with which limits. An approved tool for a firm's internal research task is not necessarily approval for a client to upload an entire case file through a personal account.

The lawyer also needs an opportunity to distinguish your statements from generated text. A summary may misstate what you said or leave out an important qualification. Preserve the underlying source and identify which parts were generated when asking counsel to review the result.

Worked example: two different ways information leaves a conversation

Fictional example: Sam receives an email from their attorney about a dispute. Sam wants simpler wording and considers pasting the whole email into a personal chatbot. The email includes names, a settlement position and the lawyer's analysis.

Before doing that, Sam can ask counsel: “May I use this particular tool to understand your explanation, and which information may I enter?” Until that is resolved, Sam can ask the lawyer directly about the confusing passage. Replacing names with initials may still leave sensitive facts identifiable.

In a second scenario, Sam's attorney proposes using an approved tool for a limited task. Sam can ask which materials it receives, who can access them and how the lawyer checks the output. The different process may matter, but this example does not pronounce either communication privileged or waived. A lawyer must analyze the actual circumstances.

The useful preparation habit is to track the proposed disclosure before it happens:

This sheet organizes the decision. It is not a consent form or an opinion about privilege.

Recording and meeting summaries raise additional issues

An AI meeting assistant may create a recording, transcript and summary from a lawyer-client conversation. Do not assume that inviting such a tool is equivalent to taking private handwritten notes.

New York City Bar Formal Opinion 2025-6 addresses AI recording and summarization, including notice, confidentiality, tool evaluation and checking summaries. It also discusses risks when clients use their own tools outside the lawyer's control.⁴ This is jurisdiction-specific ethics guidance, not a nationwide recording-consent statute or a ruling that all transcripts share the same privilege status.

Before a meeting, identify whether any participant or platform will record or generate notes. Ask the lawyer to explain the approved arrangement and any applicable consent requirements. If you need an accessibility accommodation, discuss a workable method in advance rather than silently introducing another participant or service.

If information has already been entered

Tell your lawyer what happened without creating further unnecessary disclosures. Identify the service, account, approximate date, material submitted, available settings and whether anyone else received a link or copy. Ask how to transmit that information securely.

Do not delete, alter or conceal records based on a general blog article. Preservation obligations may apply. The lawyer can assess the facts, legal consequences and appropriate response. Forwarding an old chatbot transcript to a lawyer should not be assumed to retroactively protect everything in it.

Use legal-information tools with a clear purpose

Our guide to legal information and legal advice explains the distinction between education and professional judgment. LawSensai's privacy policy describes the site's stated data practices; it is a different document from a legal opinion about privilege.

For a sensitive matter, prepare a short question list and ask an appropriately licensed attorney which channels to use. Do not describe an AI chat as privileged, attorney-reviewed or part of an engagement unless the actual circumstances support that description.

General legal information, not legal advice. Privilege, work product, confidentiality and preservation questions depend on governing law and facts. Research checked October 26, 2026.

Sources and references

1. American Bar Association, Model Rule 1.6: Comment, comments [3]–[4]. Distinguishes ethics confidentiality from evidentiary privilege and work product; model guidance is not every jurisdiction's adopted rule. ↩

2. U.S. District Court, Northern District of California, Using AI Tools in Your Case, document-upload and information-risk sections. Supports the caution about third-party storage, disclosure and privilege assumptions. ↩

3. American Bar Association, Formal Opinion 512: Generative Artificial Intelligence Tools, July 29, 2024, confidentiality section, pages 6–7. Discusses lawyer evaluation of information-disclosure risks and informed consent under the Model Rules. ↩

4. New York City Bar Association, Formal Opinion 2025-6: Ethical Issues Affecting Use of AI to Record, Transcribe, and Summarize Conversations with Clients, December 22, 2025. Supports the separate discussion of recording, client-selected tools and summary review. ↩