AI can save time. Before you paste a client email or upload a document, consider what else you are sharing.
You want help writing a client email. You copy the conversation into an AI tool, add a few instructions and get a polished response in seconds.
But what else did you share along the way?
The email may contain names, contact details, financial information or private circumstances. An attachment could include even more. Before using AI to work with client information, consider what you are putting into it.
Start with the task, then choose the information
Many everyday tasks don’t need real client details. If you want help explaining a service, drafting an appointment reminder or improving the tone of an email, a fictional example may be enough.
Instead of pasting a client’s full message, try:
“Draft a polite response to a customer asking to reschedule an appointment. Leave placeholders for their name, appointment date and contact details.”
You still get help with the writing while keeping the actual client information out of the prompt.
Removing a name alone may not be enough. A combination of location, occupation and personal circumstances could still reveal who someone is.
Know the tool you’re using
Different AI products, account types and features can handle information differently. Consider whether prompts and uploaded files are stored, used for model improvement or passed to other services through connected features.
The OAIC recommends assessing an AI product’s suitability, privacy and security risks, and who can access information entered into or generated by it. Read the OAIC guidance →
Keep your information out of AI training
Before entering business information into an AI tool, check whether your conversations and uploaded files can be used to train or improve its models. Where that use is optional, turn it off.
The controls depend on the provider, product and account type. Check the setting for the account you actually use rather than assuming every version handles information the same way.
Turning off model training is an important step, but it doesn’t necessarily stop information being stored or processed to deliver the service. It also doesn’t make it appropriate to upload every client document. Use fictional examples or placeholders whenever the task doesn’t need real client details.
Our privacy guides and short videos show you where to find the relevant controls and how to change them:
ChatGPT: Review your privacy and model-training settings→
Claude: Review your privacy and model-training settings→
Your starting rule: check the training settings before sharing business information, and keep unnecessary client information out of prompts and uploads.
Protect your account
Your AI account may contain conversations, uploaded documents and connections to other tools. Protecting access deserves the same attention as other business accounts.
Check the result before using it
A polished response can still contain mistakes. Before sending AI-assisted content to a client, check the facts, remove inappropriate detail and make sure it says what you intended.
Try this: choose one regular task and see whether AI can help using fictional information and placeholders.
For a broader review of your approach:
Download the AI Governance Checklist for Small Business→
Download the AI Security Settings Guide for Small Business→
Explore all 60 Seconds to Safety videos→
Use AI with greater confidence
Book an appointment to discuss your business’s use of AI, the information you need to protect and practical steps towards safer use. Book an appointment→
