What's in a good AI policy?
AI is becoming a permanent fixture in our lives. It’s easy to use and nearly impossible to avoid. Charities and impact focused organisations need clear policies to guide its use and governance.
Right now, 10-20% of non-profits have an AI policy but 50-70% of non-profit staff are using AI tools (the research findings are varied!). This governance gap creates organisational risk as charities move fast in practice but slow in policy, leaving staff to make decisions on their own.
At Fundsorter, we’re often asked for our view on good AI policy, and what better time to share than the week before Christmas? Here are the basic principles we recommend starting with, in case you’d like to ponder policy over the summer break.
Human verification: No AI-generated content should go out into the world without a person checking it.
Transparency: Be explicit about any significant use of AI. Be honest when AI is significantly used during a process. You likely don’t need to disclose using AI for spell-checking or internal brainstorming, but you should disclose if AI generated a report, image, or public statement.
Privacy: Use public information only. Don’t feed free generative AI tools restricted or confidential information. If you need to process this type of information, make sure you are using a paid tool with a good privacy policy and "no-training" data agreements.
Dependency: Use AI as a sounding board, helper, or support system, but don’t rely on it for high-stakes decision making.
Agreement: As a team, discuss and agree which tools you will (and won’t) use, and why.
Value: Use AI only when it genuinely adds value through greater efficiency, better quality, or expanded capability.
Continuous Learning: Keep learning and testing your views - it’s a rapidly evolving space.
Not all AI outputs are good quality.
If you’re looking for some policy documents to help you kick start your own, here are a few we rate:
Inspiring Communities has one of our favourite AI Policy docs - well thought out and locally applicable (check their website next week)
The International Committee for the Red Cross has a very robust AI policy that takes a values based, humanitarian approach
Foundation Scotland's AI policy is clear and pragmatic
The Action for Happiness AI Policy is short and light touch
EPIT’s AI Use Policy
And if you’re a funder (or advocating to a funder), this is our favourite articulation of guidance for AI use in applications: “We know that generative AI tools can be helpful when drafting and editing content.... We offer no endorsement or prohibition of these tools, but please note that generic submissions (i.e., lacking specificity in project implementation and impact) will not be considered.”