
For Australian registered tax and BAS agents. Last updated September 2026.
In short: The Tax Practitioners Board (TPB) published TPB(GS) 55/2026 on 22 July 2026, setting out how the Code of Professional Conduct applies when a registered tax agent or BAS agent uses artificial intelligence. AI use is permitted. Nothing in the Code has changed. What the TPB now expects is that you obtain client permission before entering client information into an AI tool, that you verify AI output using your own professional judgement, and that you keep records showing you did both.
If you've been using AI in your practice and quietly wondering whether you're on the right side of the rules, there's now a clear answer.
On 22 July 2026, the Tax Practitioners Board published TPB(GS) 55/2026, The use of Artificial Intelligence and the Code of Professional Conduct. It's the final version of guidance that went out for consultation in March 2026. There's also a handy two-page factsheet if you'd like the short version.
Can registered tax agents use AI in Australia?
Yes. The TPB isn't asking you to stop. The Board has said openly that it sees the productivity and client service benefits and wants practitioners to adopt AI with confidence.
Nothing in the Code has changed either. It simply explains how your existing obligations, in section 30-10 of the Tax Agent Services Act 2009 and the Tax Agent Services (Code of Professional Conduct) Determination 2024, apply when AI is part of how you work.
What does TPB(GS) 55/2026 cover?
Six parts of the Code: competence, reasonable care, confidentiality, record-keeping, professional judgement, and supervision.
The idea running through all of them is simple. AI can help with the work, but it doesn't take on any of your responsibility for it. You're still the one who must be competent. You're still accountable for the advice and its accuracy. And AI output isn't a stand-in for your own analysis.
Most practitioners we talk to already work this way. The shift is that the TPB has now outlined what it expects to see if it ever looks. The challenge we hear from advisers is that "I checked it carefully" is easier to say than to show.
Do I need client permission to put client information into an AI tool?
Usually, yes. This is the item that probably requires the most change to processes.
Code item 6 says you can't disclose information about a client's affairs to a third party without their permission, and depending on how it's set up and used, an AI tool can be that third party. Permission can come through your engagement letter or a separate signed consent. The TPB suggests letting clients know who the information goes to, where it's stored, and that AI tools may be involved. If you handle tax file number information, the Privacy (Tax File Number) Rule 2015 applies on top. The code does not definitively state that AI tools are always treated as third parties – particularly if the tool has the appropriate security frameworks.
Worth asking: when someone on your team drops a client's circumstances into a chat window, does your current engagement letter cover that?
How do I show the TPB that I checked the AI's work?
TPB(GS) 55/2026 asks for documented processes to verify AI output before you use it, records that show what you did, and a quality management system with policies you follow.
In practice, that means the tool you use should make it easy to see where an answer came from and provide a trail showing you looked. If an AI tool gives you a confident paragraph with nothing behind it, checking the work means doing the research yourself, and there's nothing to point to afterwards.
A checklist before you use AI on client work
Is this tool right for the job? Know what it does well and where it struggles.
What information am I putting in? If it relates to a client's affairs, treat it accordingly.
Do I have permission? Explained to the client and recorded.
Where does the information go and who can see it? Check storage, retention, and whether prompts are used for training the AI.
How will I check the answer? Apply your own tax knowledge and the client's specific circumstances.
What am I keeping? A record of the use, the review and the decision.
Is my team supervised? Anyone working on your behalf needs the knowledge and the oversight.
There's nothing unusual on that list. Most of it is what a careful practitioner does anyway.
How Nylon helps you meet these obligations
Nylon is an AI research tool for tax and law, used by close to 2,000 firms across Australia, New Zealand and the United Kingdom. We built it for exactly this kind of checking, and it happens to line up nicely with what the TPB is asking for.
You can see where every answer comes from. Nylon works only from trusted legal and tax sources and gives you in-text citations plus a full reference list, with links straight to the exact page of each source. Verifying an answer takes seconds instead of an afternoon, and you're left with something you can point to.
Your meetings become a proper record. The issue-spotting notetaker sits in on client calls and turns them into structured notes: the facts, the assumptions, the technical issues raised, a checklist of the legislation and rulings worth checking afterwards, and clear next steps.
It fits into the systems you already use. Integrations with SharePoint, Teams and Outlook mean the outputs land where your firm keeps things, rather than in a browser tab someone can't find in six months. Try enabling auto-save on your engagement folders to SharePoint next time you make a folder.
Security you can point to. Your client data is stored locally in Australia. Nylon also holds SOC 2 Type II and ISO 27001 certification, and the distinction matters more than the badges suggest. SOC 2 Type II tests whether controls genuinely held up across a period of time, not on one good day, and ISO 27001 requires a security management system that keeps being reviewed and re-audited. They're evidence of ongoing practice rather than a box ticked once, which is the same standard the TPB is asking you to meet with your own quality management.
Does APES 110 apply as well?
For most practitioners, yes. If you're a member of CA ANZ, CPA Australia or the IPA, APES 110 Code of Ethics for Professional Accountants is a separate code with its own AI-relevant provisions. Technology-related revisions took effect on 1 January 2025, bringing technology considerations into the professional competence and due care, and confidentiality principles. The TPB Code and APES 110 sit alongside each other, so it's worth keeping both in mind.
Frequently asked questions
When did the TPB AI guidance take effect? TPB(GS) 55/2026 was published on 22 July 2026. It doesn't create new obligations, so the underlying Code requirements it explains were already in force.
Does the TPB ban ChatGPT or other general-purpose AI tools? No. The TPB doesn't name or prohibit particular tools. What matters is whether you have client permission for any disclosure involved, whether you understand the tool's limitations, and whether you verify the output before relying on it.
Is an AI tool a "third party" under the Code? It can be, depending on how the tool is configured and used. That's why client permission and your engagement letter wording matter.
Do I have to tell clients I'm using AI? The TPB recommends telling clients about the proposed disclosure of their information, including who it goes to, where data will be stored, and whether AI tools may be used.
Who does TPB(GS) 55/2026 apply to? Registered tax agents and BAS agents providing tax agent services in Australia.
Can I keep using Nylon? Yes, Nylon is a research tool that assists accountants and advisers. Its built-in verification tools help practitioners comply with their Code obligations.
One last thought
The real gain from AI is speed on the slow, menial part of the job. Hunting down the right section, checking what's changed since you last looked, chasing a ruling back to its source: that's the work that quietly eats an afternoon, and it's what AI tools like Nylon handle well.
What you get back is time for the part that needs you. Sitting with all the facts, working out how the law applies to a particular client’s situation, and deciding what to advise.
You can hand the task to AI while still delivering value to your clients and deepening your own expertise. The thinking stays yours, and that's exactly where the value remains.

WRITTEN BY
Manraj Singh Rahi
Product Lead
Manraj leads product development and the roadmap at Nylon. With a background at a leading NZ law firm and New Zealand's Permanent Mission to the UN in New York, he brings a combination of skills to put our customers first and think about the tools advisers need most. He's also on a long-running mission to find out which fruit best describes the legal system, kiwifruit is currently leading.
More Articles

Security
Nylon is now SOC 2 Type 2 certified: what it means for your firm's data
Learn what Nylon's SOC 2 Type 2 certification means, how it compares with Type 1, and why it matters for firms using AI.
23 Jul 2026
5 min read

Product
Nylon earns landmark ICAEW accreditation for UK AI legal and tax research
Nylon is the only tool accredited by ICAEW in the Legal Information & Research Software category, following an independent technical evaluation by RSM UK.
12 Jul 2026
2 min read
