AI is already being used in Australian businesses whether there is a formal AI strategy or not.
Staff are writing emails with it, summarising documents, creating meeting notes, analysing spreadsheets and trying new AI apps as they appear.
Trying to solve that by simply saying “don’t use AI” is unlikely to work.
The better approach is to give people clear rules for using it safely.
If the business has no rules, staff will make their own
Without guidance, every employee is forced to make their own decisions.
Is a personal ChatGPT account okay?
Can customer emails be pasted into Claude?
Can an AI meeting bot join a confidential meeting?
Can an employee connect an AI app to Microsoft 365?
Can a spreadsheet containing customer names be uploaded if the employee only wants help with a formula?
When there is no company answer, people tend to choose whatever seems reasonable at the time.
That is how shadow IT grows.
AI is increasingly shadow IT with a chatbot attached.
An AI Acceptable Use Policy does not need to be enormous
A useful policy should be practical enough that staff can actually follow it.
For many businesses, it should answer questions such as:
Which AI platforms are approved?
Do staff need to use company-managed business accounts?
Are personal AI accounts allowed for business work?
What customer or personal information must not be entered?
When should information be anonymised first?
Who can approve a new AI app, browser extension or meeting bot?
Can AI tools connect to Microsoft 365, Google Workspace or the CRM?
Who reviews privacy, retention and model-training terms?
How should AI-generated information be checked before the business relies on it?
What should an employee do if they accidentally provide sensitive information to the wrong service?
The policy and the technology need to agree
There is not much value in writing “staff must use approved business AI accounts” if everyone still has unmanaged personal accounts and the business does not provide an approved alternative.
Likewise, saying “do not connect unapproved AI apps” is difficult to enforce if nobody owns the process for reviewing and approving them.
Good AI governance normally needs both:
Policy: clear expectations for staff.
Administration: managed accounts, access controls, appropriate settings, offboarding and an approval process for new tools.
Staff also need a failure path
People will make mistakes.
An employee may realise they uploaded the wrong file. A meeting bot may have joined a meeting it should not have. Someone may paste customer information into a personal AI account before remembering the policy.
The worst outcome is an employee hiding the mistake because they think they will get in trouble.
A practical AI policy should tell them exactly what to do next: stop, report it, preserve the details and let the business assess what information was involved and what response is required.
This is not about scaring people away from AI
A useful comparison is driving a car.
Cars are enormously useful. We do not avoid them because accidents are possible. We learn how to drive, follow sensible rules, maintain the vehicle and understand that some situations require more care than others.
AI should be approached in much the same way.
The answer is not to tell everybody that AI is dangerous and ban it. The answer is to understand what the tool can access, what information should be put into it and what controls the business needs around its use.
Used responsibly, AI can save time and help capable people work more effectively. Used carelessly, it can expose information the business was responsible for protecting.
AI is useful, and businesses should be looking at how to use it productively.
The point of understanding account types, privacy settings, retention and third-party apps is not to make staff afraid to touch AI. It is to give them enough information to use it responsibly.
A good policy should make the safe path easy. Give staff approved tools, explain what information can be used, provide somewhere to ask when they are unsure and make it straightforward to report a mistake.
Businesses do this with email, cloud storage, mobile phones and other technology every day. AI should be treated the same way.
This is becoming part of normal cybersecurity
Businesses already have rules around email, cloud storage, passwords, mobile devices and acceptable computer use.
AI is becoming another part of that same environment.
The Office of the Australian Information Commissioner recommends organisations conduct due diligence before using commercially available AI products and establish clear policies and procedures around the use of personal information in AI. See the OAIC’s guidance on commercially available AI products.
The goal is not to stop capable people using useful technology.
It is to stop every employee independently deciding where company information can go.
A sensible starting point
If your business has not tackled AI governance yet, start with three questions:
1. What AI tools are our staff already using?
2. Which of those accounts and services are actually controlled by the business?
3. What information are staff allowed to give them?
Those three answers will usually show you where the biggest gaps are.
If you are still working through the account side, start with Paying for ChatGPT Doesn’t Mean You Have a Business AI Account. If you are worried about plugins, meeting bots and other third-party services, read Where Does Your Data Go When You Give It to an AI App?
Q10 can help put some structure around it
Q10 can help businesses establish approved AI platforms, move users toward managed business workspaces, administer user access and settings, review connected applications and develop practical AI guidelines and Acceptable Use Policies.
We can also help integrate AI use into the business’s existing cybersecurity, incident-response and managed IT processes.
The goal is not to make AI difficult to use.
It is to make sure the business remains in control of its information while its staff get the productivity benefits.
This article provides general information rather than legal advice. AI services, privacy settings and provider terms change regularly and should be reviewed for the products your organisation actually uses.
