Cyber safety
Large Data Breaches Are Rising. What Not to Paste Into AI Tools
Keep personal and client information out of public AI tools unless your organisation has approved the tool. Share only what the task needs, check your workplace policy, and tell the right person quickly if you make a mistake. The Office of the Australian Information Commissioner (OAIC) says large-scale breaches are on the rise, so these habits matter.
What do the latest Australian data breach numbers show?
The OAIC released its latest Notifiable Data Breaches statistics on 26 October 2026. They cover January to June 2026.
There were 667 notifications in that period, compared with 662 in the previous one. The overall count barely moved. The size of the biggest breaches did.
- Breaches affecting over 1,000,000 people rose from 4 to 11.
- Breaches affecting 500,000 to 1,000,000 people rose from 0 to 3.
- Breaches affecting 100,000 to 250,000 people rose from 3 to 6.
Malicious or criminal attacks remain the leading cause. Health service providers recorded the highest number of notifications at 149, followed by education at 85. Health service providers also recorded the highest number of breaches caused by human error, at 62.
These figures are about data breaches in general. They do not say that AI tools caused any of them. They are a reminder that personal information is valuable, and that holding less of it means less can go wrong.
What does "you can't lose what you don't hold" mean for AI?
Privacy Commissioner Carly Kind said it plainly in the OAIC release. "Organisations can't lose personal information they don't hold."
The same idea applies to you. If a client's name, health details or contact information never goes into a public AI tool, it can't be exposed through that tool.
The OAIC has guidance on privacy and the use of commercially available AI products. It recommends that organisations "do not enter personal information, and particularly sensitive information" into publicly available generative AI tools. The guidance also notes that once information is entered, "it will be very difficult to track or control how it is used."
That is guidance, not a ban on every use of AI. It is a good reason to think before you paste.
What should you keep out of public AI tools at work?
Think about the people behind the information. Here are some everyday examples from Australian workplaces.
- A council officer pasting a resident's complaint, with name and address, to get a reply drafted.
- An aged care worker pasting a resident's care notes to tidy them up.
- An early childhood educator pasting a child's incident report into a chatbot.
- A medical practice receptionist pasting a patient's appointment details to write a message.
- An office worker pasting a spreadsheet of staff or client details to summarise it.
Each of these could be done another way. Describe the task without the personal details. Ask for a template reply and fill in the names yourself. Remove anything that could identify a person before you share the text.
How can you use AI tools more safely?
Start with a few simple habits.
- Check your policy. Your organisation may have rules about which tools are approved and what you can enter. If you can't find a policy, ask your manager or privacy contact before you use a public tool for work.
- Use approved tools. An approved tool may have different settings and agreements to a free public one. Your organisation decides that, not you.
- Share the minimum. The OAIC guidance says organisations should take steps to minimise the amount of personal information that is input into an AI system. Apply that to each prompt you write.
- Remove identifying details. Swap names for roles or placeholders. Leave out dates of birth, addresses, account numbers and health details.
- Check what comes back. AI output can be wrong. Read it before it goes to a client or into a record.
What should you do if you paste something you shouldn't have?
Report it fast. Do not wait to see if anything happens, and do not try to hide it.
Tell your manager, privacy officer or IT contact as soon as you notice. Tell them what you entered, which tool you used and roughly when. The sooner your organisation knows, the sooner it can decide what to do next.
The OAIC release makes the same point about breaches in general. Privacy Commissioner Carly Kind said "Timely detection and response are critical to minimising the harm caused by data breaches." Quick reporting by staff is part of that.
Mistakes happen, and a workplace that hears about them early is in a far better position than one that hears late.
What is a good next step?
Look up your workplace's AI or privacy policy this week. If there isn't one, ask who is responsible for it. Then pick one task you use AI for and rewrite your usual prompt without any personal information.
If you want a structured way to build these habits, you can try the course Responsible AI for Australian Workplaces or the AI Readiness Check.
This is general information, not legal advice. Check your organisation's policy.
Sources
Learn the habit, free
Responsible AI for Australian Workplaces teaches what never goes into an AI tool, how to ask well and how to check the answer. Short online lessons, free for individuals.