By Jamie Brennan · · 4 min read · Updated 2 August 2026

The ATO just told everyone not to ask ChatGPT. It also handed you the rule for using AI well

The ATO has warned taxpayers to ignore tax advice from ChatGPT, Claude and Gemini unless a professional checks it. It is not anti-AI. It is the clearest statement yet of the one line that decides whether AI helps your business or burns it.

Four professionals in business attire in a serious discussion around a meeting table, papers and a monitor in front of them.

The ATO has done something worth paying attention to: it named the tools. As SmartCompany reports, it is urging taxpayers to ignore tax advice from AI such as ChatGPT, Claude and Gemini, along with social-media finfluencers and unqualified sources, unless they have checked it with a registered tax professional or the ATO itself. Assistant Commissioner Anita Challen put the reason in one sentence: “Remember you are making a legal declaration that the information in your tax return is true and correct.”

A chatbot cannot make that declaration for you. You can.

Now, before this reads as another “AI is dangerous” scare, notice what the ATO actually did. It did not ban AI. It drew a line. And it happens to be exactly the line we draw for clients every week, so it is worth stealing.

The line: plumbing versus judgement

Here is the whole thing in two words. AI is brilliant at the plumbing and unreliable at the judgement.

The plumbing is the mechanical work: reading a pile of receipts, sorting transactions, summarising a long document, turning your rough notes into a tidy first draft, writing the email you have been putting off. AI does this quickly, tirelessly, and well. Let it. That is where the hours are hiding.

The judgement is the part that carries weight: what is actually true, what the law means, what is deductible, what a clause in a contract commits you to, what you are prepared to sign your name against. AI will answer all of those with total confidence, and it will sometimes be flatly, expensively wrong.

Why “confidently wrong” is the specific trap

The danger is not that AI is often wrong. It is right a lot, which is what makes it useful. The danger is that when it is wrong, it looks identical to when it is right.

A search engine hands you sources and lets you judge them. A chatbot hands you a smooth, authoritative-sounding answer with no visible seams. For low-stakes plumbing, that polish is fine. For anything you would put your name to, “sounds right” is not “is right,” and the gap between them is invisible until it costs you. That is precisely why the ATO, whose entire job is the judgement layer, said what it said.

A one-question test before you trust an AI answer

You do not need a policy document for this. You need one question: if this is wrong, who wears it?

If the answer is “nobody, it is a draft I am going to check anyway,” use AI freely and at full speed. If the answer is “me, legally or financially,” then AI can help you prepare and understand the question, but a qualified human gives the answer. Tax returns, contracts, legal obligations, anything involving safety or serious money: AI drafts, a person decides.

What this looks like inside a business

  1. Run the plumbing on AI at full speed. Drafting, summarising, sorting, first passes. This is genuine, safe productivity, and most businesses are barely using it.
  2. Put a human checkpoint on anything with consequences. A real person signs off before it is filed, sent, or promised. This is the same “keep a human on the consequential steps” point we made about treating AI connectors as systems, not switches.
  3. Write the rule down for your team, in one line: AI can draft anything, but a person approves anything legal, financial, or client-facing. That is the whole playbook, and it pairs with bringing your team’s shadow AI use into the open.
  4. Keep your experts for the judgement. Your accountant and your lawyer are not made redundant by AI, they are made faster by it. AI does their plumbing; they still make the call.

The func.digital take

The ATO just gave every Australian business a free, well-timed reminder of the single most important rule for using AI well: it is a superb assistant and a poor authority. Use it for the plumbing, keep humans on the judgement, and you get the speed without the exposure. That is not a limit to resent. It is the operating manual.

If you want help drawing that line inside your own business, working out exactly where AI should do the work and where a person needs to sign off, that is what a free digital systems audit is for. We help you get the productivity without inheriting the risk. Get in touch.

Let's talk

Book your systems audit.

A clear read on your website, tools, and automation, and the highest-impact fixes to make first. No pitch, no obligation. We reply within one business day.