Your charity AI policy
Why we have this policy
AI tools can give our people more time for the work that matters: supporting our community, looking after our supporters and delivering our services. This policy sets out how we use those tools safely, so we keep the trust of the people we work with, our supporters and our funders.
It covers everyone who does work for Your charity: staff, volunteers, board members and contractors. It applies whenever AI is used for our work, on any device, at work or at home.
How we use AI
We use AI to help with work such as drafting letters and emails, summarising long documents, tidying up data, preparing first drafts of reports and applications, and answering everyday questions.
A person is always responsible for anything that goes out. AI helps us with the work; it never makes decisions about the people we support, and it never sends anything on our behalf without someone checking it first.
Approved tools
You can use these tools for Your charity work:
- the tools you approve
Use the business versions of these tools, signed in with your work account. Business versions keep our information private and do not use it to train the AI. Free or personal accounts often do, so do not use them for our work.
To use a new AI tool, or to connect an AI tool to any of our systems (email, files, our database), ask the role that approves new tools first. They will check how the tool handles our information before saying yes.
What never goes into an AI tool
Unless a tool is on the approved list above and has been set up for that purpose, never put any of the following into it:
- names, contact details or any information about the people we support
- donor or supporter details, including gift history
- health information, case notes or anything about someone's personal circumstances
- information about staff or volunteers
- passwords, bank details or other security information
- anything we have promised to keep confidential, such as funder or partner information
The Privacy Commissioner's advice is not to put personal or confidential information into AI tools unless the provider has committed that it will not keep or share that information. If you are not sure, leave it out, or remove names and details before you start.
Checking what AI gives you
AI can sound confident and still be wrong. Before anything AI has helped with is sent, published or relied on, the person using it must check:
- Facts. Names, numbers, dates and claims are correct.
- Tone. It sounds like us and is respectful to the people it's about and for.
- Fairness. It does not stereotype or treat any group unfairly.
- Originality. It is not copying someone else's work.
Privacy comes first
The Privacy Act 2020 applies whenever we collect, use or share personal information, including when we use AI tools. Our privacy officer is your privacy officer.
Before we start any new use of AI that involves personal information, we will complete a privacy impact assessment, as the Privacy Commissioner recommends. It helps us check the use is necessary, safe and fair before anyone's information goes near it.
If we ever want to use personal information to train or improve an AI tool, we will tell people clearly when we collect their information and give them a choice about it.
Being open about AI
We are open about how we use AI. If AI has played a large part in something we send to supporters, funders or the people we work with, and it would matter to them, we say so. We never use AI to pretend to be a real person.
Funding applications
Some funders have their own rules about AI-written applications. Check each funder's guidelines before using AI on an application, and make sure everything in it is accurate and reflects our real work.
Security
- Use your work account for AI tools, with multi-factor sign-in turned on.
- Do not share AI tool logins.
- Do not install AI browser extensions or apps on work devices without approval.
- Do not connect AI tools to our email, files or database without approval from the role that approves new tools.
When something goes wrong
If personal or confidential information is put into an AI tool by mistake, or an AI tool produces something harmful, tell your privacy officer straight away. Mistakes reported early are much easier to fix, and nobody will be in trouble for speaking up.
If a privacy breach has caused, or is likely to cause, serious harm, the Privacy Act requires us to tell the Privacy Commissioner and the people affected as soon as we can. The privacy officer will manage that.
Training and review
Everyone reads this policy when they start, and again when it changes. We review it by the review date, or sooner if our tools or the law change.