Lawyers and their clients are increasingly embracing generative artificial intelligence (AI) tools, but many remain unaware of the risks of privilege waiver.
Data from the Law Society of Western Australia indicates that more than half of WA lawyers recently surveyed are using generative AI tools in practice.1 Clients and in-house legal teams are also running legal documents and advice through generative AI tools to gain efficiencies.
The rapid growth in reliance on generative AI tools is of huge assistance to both lawyers and their clients. However, increased reliance and the wealth of AI tools now available raises ethical implications for the preservation of legal professional privilege (LPP) and confidentiality.
LPP protects confidential communications and documents passing between solicitors and their clients made for the dominant purpose of legal advice or use in actual or anticipated litigation. That protection depends on the maintenance of confidentiality. If confidentiality is lost, including by disclosure to a third-party, privilege is lost, making the careful handling of privileged material essential when using generative AI. Similarly, running confidential legal documents such as deeds of settlement through generative AI tools can also create a risk of loss of confidentiality.
The risk is not merely theoretical. Many publicly available generative AI tools assert broad rights over user inputs, allowing data to be stored, processed or used for model training. Given widespread use of these tools, there is a real risk that confidential client information may be uploaded to third‑party platforms, potentially risking exposure through reuse, security vulnerabilities or data breaches, thus undermining privilege and confidentiality.
A practical response is to favour use of closed‑dataset generative AI tools, such as paid, in‑house or firm‑hosted systems, after carefully assessing how the platform handles sensitive information. These tools typically offer stronger safeguards than public platforms, but the key question remains: how ‘closed’ is closed? Practitioners should closely review the terms and conditions of any generative AI tool intended to be used to ensure confidentiality is genuinely preserved and to understand what rights the provider retains.
Judicial and regulatory guidance is beginning to reflect the concerns, although a uniform position is yet to emerge. Recent decisions of the Federal Circuit and Family Court of Australia have warned that there is a real risk that uploading documents to generative AI tools may waive privilege, urging practitioners to exercise extreme caution in doing so.2
Regulators have reinforced this position, with peak legal bodies in WA, Victoria and NSW3 collectively warning against entering confidential, sensitive or privileged client information into public AI tools. They emphasised the need to carefully review contractual terms, reflecting concerns also seen in recent UK and US decisions.
The Federal Court’s recent practice note4 commented that the risks of inadvertent disclosure of privileged material may be lower for closed generative AI tools.
The Supreme Court of Western Australia’s ‘Guidelines for the use of generative AI'5 contain a warning to litigants and practitioners that ‘the privacy and confidentiality of information and data provided to generative AI tools may not be guaranteed and the information may not be secure.’
The guidelines emphasise that caution must be exercised when employing generative AI, noting that ultimately, it is litigants and practitioners responsible for ensuring that confidential information is not disclosed to third parties.
This judicial concern is also reflected in practice, as evidenced by the Law Society of Western Australia’s survey, in which 72 per cent of those surveyed identified the risk of disclosing confidential or privileged information as a main barrier to adoption of generative AI in the legal profession.6
Key takeaway
If in doubt about a generative AI tool’s capability to maintain confidentiality, the recommended default position is not to input privileged legal advice or confidential documents into an open or public generative AI tool.
At Hall & Wilcox, we are closely monitoring evolving judicial and regulatory guidance surrounding the use of AI tools in legal settings. Penelope Ford and Lauren Separovich would be pleased to discuss strategies for safeguarding legal professional privilege and confidentiality while using generative AI.
[1] The Law Society of Western Australia, ‘Law Society of Western Australia: 2025 Use of Generative AI Survey Analysis’, 12 May 2025, page 1.
[2] Mertz & Mertz (No 3) [2025] FedCFamC1A 222; Helmold & Mariya (No 2) [2025] FedCFamC1A 163.
[3] Law Society of New South Wales, Legal Practice Board of Western Australia, Victorian Legal Services Board + Commissioner released a joint ‘Statement on the Use of Artificial Intelligence in Australian Legal Practice’, 6 December 2024.
[4] Federal Court of Australia’s ‘Use of Generative Artificial Intelligence Practice Note (GPN-AI) ’, 16 April 2026.
[5] Supreme Court of Western Australia, ‘Guidelines for the use of generative AI’, November 2025.
[6] The Law Society of Western Australia, ‘Law Society of Western Australia: 2025 Use of Generative AI Survey Analysis’, 12 May 2025, page 2.
