You signed a confidentiality agreement with your client or have a confidentialy clause in your services agreement and meant every word. Their information is safe with you. Then, mid-project, you paste their confidential brief into ChatGPT to speed things up. Helpful, harmless, done in seconds.
Except it might have just put you in breach of that very agreement. Using AI on confidential work is one of the easiest ways to break a promise you fully intended to keep. Let's make sure that doesn't happen to you.
Table of Contents
- What a confidentiality agreement or clause actually requires
- How using AI can breach an NDA in Australia
- The information you forget is confidential
- Why intention does not save you
- Who this hits hardest
- How to use AI without breaking your promise
- FAQ
- Keep your word and your work
What a Confidentiality Agreement Actually Requires
A confidentiality agreement, often called an NDA, is a promise to keep certain information private and to use it only for agreed purposes. It's one of the most common terms in client work, and it's often buried inside a wider services agreement.
The core obligation is simple. You don't disclose the client's confidential information to anyone else, and you protect it. The agreement usually doesn't list every way you might slip up. It just holds you to the outcome. So the question is never whether you meant well. It's whether the information stayed protected.
That framing matters, because it's where AI quietly causes trouble. Disclosing to a third party includes more than you'd think.
The same applies to any contract you have with a client.
How Using AI Can Breach an NDA in Australia
When you paste confidential information into an AI tool, you may be disclosing it to a third party, the company behind the tool. That can be exactly what your agreement told you not to do.
Depending on the tool and your settings, that information can be stored, processed overseas, or used to train the model, as we cover in our guide on client data and AI privacy. Once it's in, you've lost control of it. If your NDA says the information stays between you and the client, sending it into an external system can break that promise the moment you hit enter, whether or not anything bad ever comes of it.
A breach of confidentiality isn't about harm. It's about the disclosure itself. The exposure exists even if nobody ever sees the data again.
The Information You Forget Is Confidential
Confidential information is broader than trade secrets and passwords. It's easy to underestimate, which is how good people slip up.
It can include the client's business plans, customer lists, pricing, unreleased products, internal documents, and the contents of the brief itself. Plenty of everyday material you'd happily feed to AI for a quick summary is confidential under your agreement. The draft you want polished, the strategy you want sharpened, the notes you want tidied. All of it can be covered.
The fix starts with a mindset shift. Before anything client-related goes into AI, assume it might be confidential and check, rather than assuming it's fine because it feels mundane.
Why Intention Does Not Save You
This is the hard truth. You can breach a confidentiality agreement with the best intentions in the world.
You weren't being careless. You were trying to do better work, faster, for the client. None of that changes the legal position. The agreement asks whether the information was kept confidential, not whether your heart was in the right place. A breach is a breach, and the consequences, from a damaged relationship to a claim for loss, don't soften because you meant well.
That's why this is worth taking seriously now, before it happens. The clients who'd care most about a breach are exactly the ones who made you sign the agreement in the first place. For a plain-English overview of contracts and confidentiality, business.gov.au is worth a look.
Who This Hits Hardest
Some businesses live in confidential information all day, which puts them most at risk.
Virtual assistants and online business managers handle their clients' inboxes, files, and systems. Consultants and coaches hear sensitive details about businesses and people. Designers and agencies see unreleased products and strategies. If that's you, AI is a daily temptation and a daily risk at the same time. The more confidential material passes through your hands, the more a single careless paste can cost you.
The answer isn't to avoid AI while your competitors race ahead. It's to set up clear rules and the right contract so you can use it safely.
How to Use AI Without Breaking Your Promise
-
Check the agreement first. Know what you've promised before using AI on any client work.
-
Strip identifying and confidential details before pasting. Use placeholders for the sensitive parts.
-
Turn off training and data retention in your AI tool's settings.
-
Ask the client if you want to use AI on genuinely sensitive material. Permission changes everything.
-
Use a services agreement that addresses AI and confidentiality clearly for both sides.
FAQ
Can using ChatGPT actually breach an NDA?
It can. Pasting confidential information into an AI tool may count as disclosing it to a third party, which is often exactly what an NDA prohibits. Check your agreement and strip sensitive details.
What if I delete the chat afterwards?
Deleting the chat doesn't undo the disclosure. Depending on the tool, the information may already be stored or processed. The safer move is to not put confidential material in to begin with.
How do I protect myself and still use AI?
Strip confidential details, adjust your tool's data settings, get client permission for sensitive work, and use a contract that sets clear rules on AI and confidentiality.
Keep Your Word and Your Work
Confidentiality is the foundation of client trust, and AI can crack it without you noticing. You don't have to choose between working efficiently and keeping your promises. You just need clear habits and the right paperwork.
A solid Services Agreement sets out confidentiality on both sides and can address how AI fits in, so you and your client know exactly where the lines are. Pair it with sensible habits and you can use AI without ever breaking your word. For the privacy side of the same problem, read our guide on client data and AI.
Protect your client relationships before your next project.
About the Author

Riz is the Founder & Director of Foundd Legal, a lawyer with 20+ years' experience and a long history of building online and ecommerce businesses.
She helps creatives and online business owners protect and grow their businesses with clear, practical legal tools that actually make sense.
SIGN UP TO OUR FREE BUSINESS CHECKLIST
Disclaimer
We do our best to keep this content accurate and up to date, but laws change, interpretations evolve, and the internet isn't perfect. Occasionally, information may be outdated or contain errors.
This content is for general information only and isn't legal advice. If you choose to rely on it, you do so at your own discretion. For advice specific to your business, you'll need support tailored to your situation.
All rights reserved. © Foundd Legal Pty Ltd