Law

QLS Guidance Statement No. 37: What Brisbane Solicitors Need to Know

Published 17 August 2026

Short answer: QLS Guidance Statement No. 37 doesn't ban AI. It confirms that a solicitor's existing duties of confidentiality, competence, and independent judgment apply in full to AI-assisted work, that AI output must be verified by someone competent to check it, and that supervisors and legal practitioner directors are responsible for firm-wide AI systems. A documented firm AI policy is the recommended starting point.

Queensland Law Society was the first law society in Australia to issue formal guidance on AI in legal practice: Guidance Statement No. 37, Artificial Intelligence in Legal Practice. It sits alongside a newer national companion document, the AI Selection and Use Checklist, published in February 2026. It's worth keeping those two straight: GS37 itself is the older of the two, and the Checklist is the newer national layer on top of it.

Neither document bans AI use. Both make clear that solicitors are personally accountable for how they use it, and that the ordinary duties of practice don't get suspended just because a task was AI-assisted.

What GS37 actually requires

GS37 points to ASCR Rules 4, 5, 9, 17, 19 and 37 as directly relevant to AI use. In practice, that breaks down to:

  • Confidentiality and fiduciary duty, especially relevant the moment any third party, including an AI vendor, gets access to client data.
  • Competence and independent judgment. You're still responsible for the output. AI is predictive, not authoritative, and it can be confidently wrong.
  • Verification, not delegation. AI-assisted work must be checked by someone competent to know if it's wrong before it goes anywhere near a client or a court.
  • Firm-level responsibility. QLS points to supervisors and legal practitioner directors as accountable for firm-wide systems, not just the individual solicitor who typed the prompt. A documented, communicated AI policy is the recommended starting point.

The part that makes this different from a normal compliance question

Here's where legal practice diverges from, say, accounting: client consent alone doesn't fully protect you. A US court has already found that running privileged documents through a third-party AI tool can waive privilege, not because the client didn't know, but because privilege depends on confidentiality being maintained in substance. Once privileged material has touched an external system, a court can find that substance was broken, regardless of what the client agreed to.

Australian courts haven't ruled on this yet. That's exactly why firms are treating it as a live risk instead of a settled question. Nobody wants to find out the hard way, in a matter that actually matters.

What this means practically

A signed client authority to "use AI tools" is good practice and worth having. But for privilege specifically, the more important question is what the AI tool itself does with the data: does it train on inputs, does a vendor retain or access it, is there a real third party in the data path at all? Genuine due diligence on the tool matters more here than in most other compliance contexts.

Firms that can say, plainly, "our AI tools run entirely on infrastructure we own and control" have a materially stronger position, both for the ethics question GS37 raises and for the privilege question no Australian court has answered yet.

Want a second opinion on your firm's AI setup?

Book a confidentiality review, no obligation.

Book a Consultation