TPB guidance on AI: What accountants and bookkeepers in Australia need to know

Published Monday 21 September 2026

Table of contents

• What is TPB(GS) 55/2026?
• Who does TPB(GS) 55/2026 apply to?
• What are the key TPB obligations?
• What you need to do and what it looks like in practice
• How Xero supports you

Key takeaways

• TPB(GS) 55/2026 clarifies how AI use intersects with your existing TASA obligations – no new requirements, just guidance on applying current rules when AI is involved.
• Three core obligations define TPB expectations: understand AI's limits and biases, maintain human control and review of all outputs, and document your AI workflows for quality management and compliance.
• Client consent is mandatory before using AI with their data; you must disclose which tools you use, how information is shared, where data is stored, and give clients a clear choice to opt out.
• Tax practitioners remain fully accountable for the accuracy and appropriateness of advice and information – AI outputs should support, not replace, your professional judgment.

What is TPB(GS) 55/2026?

The Tax Practitioners Board Guidance Statement 55/2026 helps accounting professionals understand how using AI impacts their existing legal obligations as tax agents. The statement looks at the relationship between AI use and the Tax Agent Services Act 2009 (TASA), including the Code of Professional Conduct (section 30-10 of the TASA).

The statement doesn’t set out any additional rights or obligations that aren’t already covered in the TASA – it simply provides guidance on how existing obligations should be considered when a tax professional uses AI. As the use of AI increases across accounting and business advisory, this guidance statement is a starting point that helps all Tax and BAS agents meet their obligations while keeping their clients safe. If you’re unsure about what you need to do, it’s best to refer back to the Tax Practitioners Board or seek independent legal advice to suit your practice.

Who does TPB(GS) 55/2026 apply to?

Registered tax agents and BAS agents (which the TPB collectively calls ‘tax practitioners’) need to follow the guidance in this statement. Specifically, when using AI in tax services.

Tax practitioners will already be following their statutory obligations under the TASA. But, it’s time to consider how compliance with these obligations differs when AI is brought into the picture. We take a closer look at specific obligations in relation to AI use below.

What are the key TPB obligations?

The guidance statement draws out several Code and Determination items from the TASA that have an impact on AI use. Based on those items, the requirements for tax practitioners using AI broadly fall into the following categories:

• Understand it. Tax practitioners need a clear view of what AI does in their practice, its purpose, and its limits. AI has biases and complexities that are hard for humans to understand, and it can hallucinate inaccurate information – tax practitioners should take this into consideration when they apply the technology. When AI tools are used for tasks outside of inhouse expertise, the outputs should be verified by a person with specialist knowledge.
• Control it. AI outputs should be reviewed, verified, corrected or rejected by a human before they’re relied on. Tax practitioners should establish processes for reviewing and interpreting AI decisions and outputs. AI outputs shouldn’t be used as a substitute for professional judgement, or analysis of a client’s circumstances – tax practitioners remain accountable for the accuracy of information and advice provided to clients.
• Document it. Tax practitioners should document how and when AI generated content, decisions, or analysis is reviewed and verified. To support quality management in your practice, policies and procedures you use to manage AI must be documented and enforced with a clear trail of evidence.

Alongside these three key areas, there’s confidentiality to consider. Tax practitioners need consent from clients before sharing information relating to their affairs with third parties. Under the TASA, AI models and tools can count as third parties, so practitioners must get permission from clients before using them. Be sure to give clients a clear and comprehensive explanation about how the AI will be used, which information will be shared, the specific tools, and where data is stored. Getting permission can be as simple as a signed letter of engagement, or another form of signed consent.

AI use also has implications for The Privacy Act 1988. Commercial AI tools, or ones created or modified by practices, need to be checked regularly to ensure data is secure.

What you need to do and what it looks like in practice

1. Review AI outputs and decisions with your own judgment before accepting them

• Check transactions that have been auto-reconciled by AI software to make sure matches and tax codes are correct.
• Review draft communications generated by AI tools before they're sent to clients.

2. Get expert verification when you lack in-house knowledge

• For instance, you should not use AI to sell services that you aren’t qualified to provide and you should always referee work to appropriately qualified people, such as auditors or financial planners, if you don't have the skills in-house.

3. Document and enforce clear policies and review procedures for all AI output

• Establish and regularly review your processes for checking and verifying AI outputs to ensure there's a clear data trail.
• Update these procedures as your technology use evolves.
• Supervise your staff and AI tools.
• Document the processes your staff use, as well as the AI-generated outputs, so you can fully trace practice workflows.

4. Get explicit client permission before using AI

• Update letters of engagement to explain what AI tools you use, how they use and store information, and when they'll be involved.
• Provide clients with a clear place to refuse or confirm consent.

5. Keep accurate records and quality management systems

• Maintain and assess your client records and quality management systems regularly, particularly as your AI use evolves.
• Review any insights an AI tool surfaces about your clients' financial picture, contextualizing and explaining them before sharing.

6. Conduct due diligence on your AI tools

• Check your tools against TASA obligations, the Privacy Act, and your own policies.

How Xero supports you

AI features in Xero are governed by our Accountable Intelligence approach. This approach is about partnering AI tools and human insight, so that AI processes and decisions are visible and explainable, and you always have full oversight and the final say.

There are four parts to Accountable Intelligence:

• Transparent: Our smart AI companion, JAX (Just Ask Xero), shows its work behind every answer it provides you. You can see the data, and get a visible audit trail that explains exactly how a conclusion was reached.
• You stay in control: You can review, approve, or override any action. When JAX auto-reconciles statement lines, you can head over to the ‘Reconciled’ page for a full view of JAX’s work, making corrections and checking supporting documents, or simply reviewing the transaction history.
• Grounded in real data: JAX Assure, Xero’s proprietary control system, supports increased accuracy and fewer hallucinations compared to conventional AI tools that rely solely on Large Language Models.
• Your data stays yours: Data isn’t used to train third-party LLMs. Where Xero uses LLMs, they’re fully vetted, just as any other data subprocessor would be by Xero.

Xero’s AI features keep you in control

The TPB’s response to AI use in the accounting industry shows just how quickly the technology is impacting tax practitioners. Providing that practices understand, control, and evidence AI use, the technology can make a huge difference for practices, enabling more efficient, accurate, and personalised services for clients.

Matching AI tools with human judgement is the best way to handle the technology in your practice. Xero’s AI features keep you in control, so you get the best bits of AI – accuracy, efficiency, insights – paired with your earned experience, skills, and judgement.