What can I do to set myself up for success when using AI tools?
Verify the AI tool within your entity
For example, before using an AI tool, you could consult your IT Helpdesk, or your entity may have a list of permitted AI tools with an outline of permitted uses of various AI tools.
- Your entity may deem certain AI tools low risk for use with OFFICIAL Government and/or OFFICIAL Sensitive information due to security controls in place.
- Other tools may be deemed as high risk for use with OFFICIAL Government information and/or OFFICIAL Sensitive information as the data could be made available to the public either through model training, data breaches on stored data, sharing with third parties, or storage and processing outside of Australia. This is often the case with no or low-cost web-based publicly available AI tools such as ChatGPT, Claude and Gemini.
- Note: DeepSeek is currently restricted for WA Public Sector Staff.
Do not enter OFFICIAL Government information not already publicly available online into a tool that has not been assessed by your IT Helpdesk
If the tool has not been assessed by your IT Helpdesk, do not enter any government information that is not already publicly available online. For additional security use any “opt-out” features for large language models that may retain and use information for training purposes. Disable training and logging features, as well as chat history where possible.
Review and verify outputs for quality and accuracy
AI-generated content can sometimes be unreliable and inaccurate. You should always critically assess all AI generated outputs by checking their quality and accuracy. Pay attention to the wording and sequence of your prompts. AI can easily provide different outputs based on what you enter.
You should not use AI tools to replace your own research, analysis and content development, or ‘copy and paste’ AI generated content into your work without appropriate review. Ensure you can explain your content and reference your research – you are responsible for the content you develop.
Be aware of potential bias and nuance
AI models are trained to make predictions given a specific context and can therefore lack real-world understanding, providing misleading outputs at times. AI outputs may also inadvertently reinforce stereotypes or inequalities.
To manage this risk, you can learn about diversity and inclusion to improve your ability to identify and critically evaluate unfair, biased, non-inclusive, or discriminatory representations. Think about the data that may have been used to train the tool.
Be transparent with your outputs
When publicly available content has been created using an AI tool, you should consider including a disclaimer such as in the example below:
“Note: AI was used to create this image.”
Practical scenarios
The following scenarios are examples of how you can safely and responsibly use AI for your work, including the use of non-secure tools. Using AI will depend on the task, sensitivity of the information and your agency’s policies.
Summarising complex information
Show moreHow do you use AI?
You have heard of a generative AI tool called NotebookLM and check in with IT Helpdesk at your entity to ensure you can use this tool. IT Helpdesk reminds you to only enter publicly available or unofficial information into this tool.
You upload a publicly available policy into NotebookLM and ask for a summary of the document.
You interrogate the AI further by asking for specific information on areas of interest you want to know more about in the policy.
Knowing that generative AI can produce inaccurate outputs, you independently check any information generated by NotebookLM which you plan to use in drafting your advice.
Assisting with initial research
Show moreHow do you use AI?
You are keen to use the free version of ChatGPT and check with IT Helpdesk on the permitted use of this tool.
You are advised that you can only enter UNOFFICIAL information or publicly available OFFICIAL Government information into the tool.
You use the chat function in ChatGPT to ask basic questions regarding the technical field and request it to explain key aspects of the field in simple terms.
Upon the AI generating a few responses, you ask questions and request definitions for technical terms you have encountered.
Now that you have a basic understanding, you ask to have a discussion with a subject matter expert in the team to ensure your understanding is correct.
Content creation
Show moreHow do you use AI?
You check your entity’s AI tools reference list and note that Microsoft 365 CoPilot can be used with OFFICIAL Government information as long as you are logged into your work email address and the green shield that denotes enterprise protection is displayed.
You explain to the tool who the target audience is and the information you would like to convey in an interesting manner to maximise engagement with the presentation.
You begin brainstorming with the AI, inputting your presentation text.
The AI outputs several titles and changes to your presentation’s text. You consider and maintain the integrity of the information you are trying to convey.
Upon being satisfied with many of the suggested changes, you implement them when drafting your presentation.
Before adopting or procuring new AI tools, please check they are compliant to your entity’s requirements.
Print this page
