Accounting

What TPB(GS) 55/2026 Actually Means for Your Accounting Practice

Published 6 August 2026

Short answer: TPB(GS) 55/2026, finalised by the Tax Practitioners Board on 22 July 2026, confirms that entering a client's information into an AI tool can count as a disclosure to a third party under Code Item 6 of the Code of Professional Conduct. If your practice uses AI with client data and doesn't have specific, documented client permission for it, you may already be in breach.

On 22 July 2026, the Tax Practitioners Board finalised TPB(GS) 55/2026, its first formal guidance on how the Code of Professional Conduct applies when a tax or BAS agent uses AI. If you haven't read it yet, here's the short version: entering a client's information into an AI tool can itself count as a disclosure to a third party. If you don't have documented permission for that, you may already be in breach.

This isn't a hypothetical compliance risk sitting in a PDF nobody reads. It's live, dated guidance that most Brisbane practices haven't caught up with yet.

The core issue: Code Item 6

Code Item 6 of the Code of Professional Conduct has always said you can't disclose information about a client's affairs to a third party without their permission, unless you're legally required to. That's not new. What's new is the TPB confirming, in writing, that an AI tool counts as a third party for this purpose.

Practically, that means: if a staff member pastes a client's bank statement, trust deed, or working papers into a generic AI chatbot, and the client hasn't given documented permission for that specific kind of disclosure, it's treated the same way unauthorised outsourcing or offshoring would be.

What the guidance actually requires

  • Documented, specific permission, not a generic privacy policy clause. The TPB wants clients told what's being disclosed, to whom, where it's stored, and whether AI tools may be used at all. That permission needs to be captured somewhere reproducible: a signed engagement letter, signed consent, or general authority.
  • Ongoing accountability for AI output. You can't treat AI as a black box. Practitioners are expected to review and verify what AI produces, understand roughly what the tool does with inputs, and be able to show that review happened.
  • Awareness of embedded AI. This is the part catching firms off guard. Platforms like Xero, MYOB, and Microsoft 365 increasingly have AI features switched on by default. If those features process client data, that's a disclosure event too, even though nobody in the firm consciously decided to "use AI" that day.

Does this apply to BAS agents too?

Yes. The Code of Professional Conduct under the Tax Agent Services Act covers both tax agents and BAS agents, and the same considerations apply to both.

What to actually do about it

Two things need to happen, and they're not mutually exclusive:

  • Update engagement letters to capture specific, informed consent for AI use, naming the categories of tool, where data goes, and how it's stored.
  • Audit your stack for AI features you didn't consciously turn on, so you're not disclosing client data through a channel you forgot existed.

There's a third option worth knowing about too: if the AI itself never sends client data to a third party, because it runs entirely on hardware inside your own office, the disclosure question changes shape. You still need proper consent on file, but the practical answer to "where does it go" becomes "nowhere, it stays here," which is a much easier sentence to put in an engagement letter and mean it.

Not sure where your practice stands?

Book a 20-minute AI and confidentiality audit, no obligation.

Book a 20-Minute Audit