Before You Pick a Tool, Check the Rules
In Level 1 we looked at what AI can and can't do for your charity. Now for the practical question: which tool should you actually use?
It's tempting to open whatever chatbot is already on your phone. But your phone is your own space and work belongs to the charity. Think of it like the minibus. You might be a brilliant driver, but you still check that you're insured and allowed to drive it before you set off.
So in this lesson we'll do three things. We'll find your policy, compare the account types and then pick a tool you're allowed to use.
Step One: Find Your Policy
Does your organisation have an AI policy? Lots of smaller charities don't have one yet. That's fine, and it's normal.
Start by looking in the obvious places: the staff handbook, the shared drive, the intranet and your volunteer agreement. Search for "AI", "artificial intelligence", "acceptable use" or "IT policy". Rules on AI are often tucked inside an older IT or data protection policy.
Once you find it, sum it up in three lines. Which tools are approved? What must never go into them? Who do you ask when you're unsure?
What if there's nothing at all? Then your job is to name the person you'd ask. That could be your line manager, your IT contact, your data protection lead or a trustee. Send a short email: "I'd like to use an AI assistant for drafting. Which tools are we allowed to use?" Asking first protects you and the charity.
Step Two: Personal Accounts vs Work Accounts
Here's the big idea for today. The same brand of tool can handle your data very differently depending on the kind of account you're signed into.
A personal account is like chatting in a café. A work account is more like a meeting room the charity has booked, where the rules are set out in a contract. Let's look at what each vendor's own help pages said in October 2026.
ChatGPT. On the Free, Plus and Pro plans, OpenAI can use your chats to train its models by default. You can switch that off under Settings, then Data Controls, then "Improve the model for everyone". On ChatGPT Business and Enterprise, OpenAI says it doesn't train on your inputs and outputs by default.
Claude. Since Anthropic changed its consumer terms in 2025, people on the Free, Pro and Max plans choose whether their chats can be used for training. If you say yes, new chats can be kept for up to five years. If you say no, the usual 30-day retention applies. None of this covers the work plans, Team and Enterprise.
Gemini. With a personal Google account, a setting called "Keep Activity" controls what's saved. Even with it switched off, Google still keeps your chats for 72 hours. Some chats are read by human reviewers, and those can be kept for up to three years. Google's own advice is not to enter confidential information you wouldn't want a reviewer to see. With a work Google Workspace account, Google says your content isn't reviewed by humans or used to train models outside your organisation without permission.
Microsoft 365 Copilot. If you sign in to Copilot Chat with your work Microsoft account (Microsoft calls it an Entra account), you get "enterprise data protection". Microsoft says your prompts and responses then aren't used to train its foundation models.
Do you see the pattern? Work accounts usually come with stronger promises written into a contract. Personal accounts lean on settings that you have to find and check yourself.
One honest caution: these settings change. Always read the vendor's current help page instead of trusting a blog post, or this lesson, a year from now.
Step Three: Check the Charity Deals
Good news. Several vendors have charity programmes, and some are generous. Here's what they offered in October 2026.
- Google Workspace for Nonprofits is free for eligible organisations and now includes the Gemini app and NotebookLM with work-level data protections. Paid tiers add Gemini inside Gmail and Docs.
- OpenAI for Nonprofits offers ChatGPT Business from $8 per user per month, billed yearly, and up to 75% off Enterprise. Eligibility is checked through Goodstack.
- Claude for Nonprofits, launched in December 2025, gives up to 75% off the Team and Enterprise plans and includes a free course called AI Fluency for Nonprofits.
- Microsoft has offered eligible nonprofits a 15% discount on Microsoft 365 Copilot since November 2024. Copilot Chat with enterprise data protection is also available to anyone signed in with a work account.
Prices are in US dollars and offers come and go, so check the current page before you rely on any of them.
Putting It Together: Priya's Choice
Let's meet Priya, a volunteer coordinator at a food bank. Her charity already runs Google Workspace for Nonprofits. Her manager confirms that Gemini, used through her work account, is approved for drafting rotas and volunteer emails.
Priya writes a short note for her own records: "Tool: Gemini via our work Google account. Why: approved by my manager on 20 October; work account means chats aren't human reviewed or used for training outside our organisation; no extra cost under Google for Nonprofits. Also checked: ChatGPT Business nonprofit pricing, not needed."
That's it. It takes five minutes, and now she can explain her choice to anyone who asks.
Your Turn
Try this on your own job this week:
- Find your AI or IT policy, or name the person you'll ask.
- Open the help page for one tool and note its training and retention settings for both personal and work accounts.
- Check whether your charity qualifies for a discount or free option.
- Write a three-line note like Priya's: the tool, why you chose it and what else you checked.
In the next level we'll look closely at what must never go into a prompt, even in an approved tool.
Key Takeaways
- Check your policy first. If there isn't one, ask a named person before you start.
- Personal and work accounts of the same tool can handle your data very differently.
- Read the vendor's current help pages, because settings and offers change.
- Google, OpenAI, Anthropic and Microsoft all offered charity deals or free options in October 2026.
- Write down which tool you chose and why.