Legal privilege in the era of Generative AI

The rise of Generative Artificial Intelligence (“AI”) has fundamentally altered how individuals seek information, including legal guidance. Millions of users across the globe now routinely consult AI chatbots such as OpenAI’s ChatGPT or Anthropic’s Claude, for assistance on issues ranging from employment disputes and commercial transactions to criminal defence strategies.

The intuitive and conversational interface of these tools creates a deceptive sense of confidentiality. However, a landmark 2026 ruling from the United States District Court has delivered a stark warning: conversations with AI chatbots are not shielded by attorney-client privilege. 

This raises an important question in the Indian context: are such communications protected under Indian law? This article examines the contours of attorney-client privilege in India and evaluates whether interactions with AI chatbots fall within its scope.

In United States v. Heppner, a United States District Court for the Southern District of New York, in a criminal prosecution involving securities fraud, wire fraud, and conspiracy, held that written exchanges between Bradley Heppner, former CEO of GWG Holdings, Inc., and Anthropic’s generative AI platform Claude were neither protected by attorney-client privilege nor the work-product doctrine.

Three aspects of the ruling are particularly instructive. First, the Court held that Claude is not an attorney. Second, the Court found a complete absence of confidentiality in light of Anthropic’s privacy policy, which permits data collection, model training on user inputs, and disclosure to third parties, including government agencies. Third, the Court highlighted a collateral risk: Heppner had input his lawyers’ advice into Claude, which could have compelled his defence counsel to testify regarding otherwise privileged communications

The ruling closed with an observation that resonates across jurisdictions: “AI’s novelty does not mean that its use is not subject to longstanding legal principles.”

While Indian law recognises various forms of privilege, such as spousal privileges, only the attorney–client privilege is relevant in the present context. The foundation of attorney-client privilege in India lies in Sections 132 to 134 of the Bharatiya Sakshya Adhiniyam, 2023 (“BSA”) (previously Sections 126 to 129 of the Indian Evidence Act, 1872 (“Evidence Act”)), which codify common law principles governing professional communications.

Section 132 of the BSA (previously Section 126 of the Indian Evidence Act, 1872) applies exclusively to advocates enrolled with a State Bar Council under Section 29 of the Advocates Act, 1961. It prohibits an advocate from disclosing, without the client’s express consent, any communication made in the course of professional employment, the contents of any document accessed in that capacity, or any advice given to the client. This protection extends to communications made through third parties on the client’s behalf and applies equally to interpreters, clerks, and employees of advocates.

Section 134 of the BSA further stipulates that no person can be compelled to disclose confidential communications with their legal advisor, unless they voluntarily choose to appear as a witness. The term “legal advisor” is not limited to enrolled advocates, it extends to any individual who acts in a legal advisory capacity to a client, including in-house counsel advising their employer. Communications made in such a context are therefore protected by the rule of non-compellability under Section 134.

Section 227 of the Companies Act, 2013, protects privileged communications made to legal advisors from disclosure before the National Company Law Tribunal, the Central Government, the Registrar, or an inspector appointed by the Central Government.

Similarly, Order XI and Section 30 of the Code of Civil Procedure, 1908 that empower courts to order discovery and production of documents exclude privileged material from discovery, in camera. Even in criminal proceedings, where Section 91 of the Code of Criminal Procedure, 1973 (now Section 94 of Bharatiya Nagarik Suraksha Sanhita, 2023 (“BNSS”)) which authorises the production of any document before a court, exclude privileged communications between a lawyer and client.

Indian courts have clarified that privilege arises only upon the formation of a formal attorney-client relationship. In Kalikumar Pal v. Rajkumar Pal (1931), the Calcutta High Court held that privilege does not extend to communications made prior to such a relationship.

Further, in Memon Hajee Haroon Mohamed v. Abdul Karim (1878), the Bombay High Court emphasised that privilege under Section 126 of the Evidence Act (Section 132 of BNSS) applies only to communications that are inherently confidential. The Calcutta High Court in Bakaulla Mollah v. Debiruddi Mollah reaffirmed that the privileged communications disclosed by the attorneys in the course and for the purpose of his employment, remain inadmissible in evidence.

First, and foremost, no privilege can ever arise between a user and AI Chatbots as they are not enrolled advocates under the Advocates Act and do not owe enforceable professional duties. The AI chatbots also do not qualify as legal advisors in law. As established in Kalikumar Pal, privilege cannot arise in the absence of a formal attorney-client relationship. In the case of an AI chatbot, no such relationship can be formed, and consequently, no privilege can arise.

Second, confidentiality between the user and an AI chatbot is entirely absent. Generative AI platforms process user inputs on cloud servers operated by corporations, often in foreign jurisdictions. Their terms of service permit data storage, model training on user inputs, and disclosure to third parties. As noted in Heppner, a user who voluntarily communicates through such a platform cannot claim a reasonable expectation of confidentiality. Therefore, the requirement of a “confidential communication,” as articulated in Memon Hajee Haroon Mohamed, is not satisfied.

Third, privilege under Section 132 of the BSA attaches specifically to legal advice provided “in the course and for the purpose of such employment.” AI tools expressly disclaim providing legal advice; instead their terms of use often direct users to consult qualified professionals. Therefore, a user cannot credibly claim to have sought legal advice from a tool that has expressly disclaimed providing it.

Fourth, generative AI platforms qualify as intermediaries under Section 2(w) of the Information Technology Act, 2000 (“IT Act”). Under Rule 3(1) of the Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Rules, 2021 (“Intermediary Guidelines”), they are required, in order to maintain safe harbour protection, to preserve user data and provide it to lawfully authorised government agencies for the investigation or prosecution of offences, or to courts in civil and criminal proceedings.

Consequently, investigating agencies, such as the Enforcement Directorate under Section 50 of the Prevention of Money Laundering Act, 2002, the Narcotics Control Bureau under Section 67 of the Narcotic Drugs and Psychotropic Substances Act, 1985, the Serious Fraud Investigation Office under Section 217 of the Companies Act, 2013, or an officer in charge of a police station under Section 94 of the BNSS, as well as courts, may compel such platforms to disclose user interactions, including discussions relating to legal strategy, admissions of fact, or evidentiary matters.

In sum, communications with AI chatbots in the context of legal advice or strategy cannot attract legal privilege, nor is any procedural mechanism available to assert or enforce such protection. Further, where a user discloses or uploads privileged advice previously obtained from their lawyer to an AI chatbot, such conduct may constitute an implied waiver of privilege, thereby jeopardising the confidentiality of the original communication.

The risks are particularly acute in the corporate context. Indian companies under regulatory scrutiny, whether from the Securities and Exchange Board of India, the Competition Commission of India, or with respect to prosecution by the Enforcement Directorate against senior management, must remain conscious that AI-mediated communications relating to legal strategy attract no form of legal privilege. The use of AI tools to evaluate litigation risk, prepare confidential internal notes, or assess regulatory exposure may therefore expose such material to disclosure in the course of investigations or discovery proceedings.

The practical implications are clear. AI chatbots are no substitute for a privileged attorney-client consultation. That said, certain safeguards should be observed when using AI tools for issue-spotting, document drafting, or the preliminary assessment of legal strategy:

• All sensitive and identifying information must be masked, anonymised, or replaced with hypothetical scenarios before being inputted into any AI platform.

• Documents containing privileged material, including legal memoranda, counsel opinions, investigation reports, and correspondence with advocates, must not be uploaded to generative AI platforms, as doing so constitutes disclosure to a third party and may result in the loss of privilege.

• AI-powered transcription and note-taking services are equally unprotected.

• Legal consultations, strategy discussions, and communications with counsel must not be recorded using such tools; any transcripts generated carry no privilege and may be subject to compelled disclosure.

About the author: Prabudh Singh is a Senior Associate at AP & Partners.

Disclaimer: The opinions expressed in this article are those of the author(s). The opinions presented do not necessarily reflect the views of Bar & Bench.

If you would like your Deals, Columns, Press Releases to be published on Bar & Bench, please fill in the form available here.