The Use of Generative AI and Risks to Legal Professional Privilege

The implications of generative artificial intelligence (AI) in legal proceedings are increasingly coming under judicial scrutiny, particularly in relation to confidentiality and legal professional privilege.

Courts in New South Wales have adopted a cautious approach to the use of generative AI in documents prepared for court. These concerns were articulated by Bell CJ in his address to the Australian Bar Association in August 2025, where his Honour highlighted risks associated with confidentiality, privacy and the preservation of legal professional privilege when AI tools are used in legal practice.

Recent US Decision on AI and Privilege

These issues have recently been considered by a United States District Court in circumstances where legal privilege was claimed over documents described as an AI‑generated outline of a defendant’s defence strategy.

In United States v Heppner (No. 25‑cr‑00503‑JSR, S.D.N.Y.), the Court ruled that documents created by a party using a commercial generative AI platform were not protected by legal privilege.

Although the decision is not binding in Australia, the reasoning is highly relevant. In Australia, legal professional privilege protects confidential communications and documents created for the dominant purpose of obtaining or providing legal advice, or for use in existing or anticipated legal proceedings.

Why Privilege Was Not Protected

The US District Court found that privilege did not apply for two key reasons.

First, the documents uploaded to the AI platform were not communications between the defendant and his lawyers. Instead, they were inputs provided to a third‑party technology platform, which broke the necessary lawyer–client relationship required to attract privilege.

Secondly, the Court found that the documents were not confidential. This conclusion was based on the AI platform’s privacy policy, which users are required to accept. The policy stated that:

• the platform collects data relating to users’ inputs and generated outputs;
• that data may be used to train and improve the AI system; and
• the data may be disclosed to third parties.

Because the information was shared with, and capable of being used by, the AI provider and others, the Court held that the documents lacked the confidentiality necessary for legal privilege to apply.

Key Takeaways for Australian Litigants and Practitioners

While the decision arises from the United States, it serves as a clear cautionary tale for Australian litigants and legal practitioners.

Uploading sensitive material, draft legal strategies or factual instructions into commercial generative AI platforms may risk:

• waiver or loss of legal professional privilege;
• unintended disclosure of confidential information; and
• difficulty relying on AI‑generated outputs in court proceedings.

As Australian courts continue to examine the role of generative AI in litigation, parties should exercise significant caution when using these tools, particularly where confidential or privileged information is involved. Careful consideration should be given to the terms of use and privacy policies of AI platforms, and to whether alternative, secure methods are available.

The evolving judicial approach suggests that, without appropriate safeguards, reliance on generative AI may expose parties to significant legal and evidentiary risks.