Charity AI safeguarding reviews fell from 36% to 19%. Here's why that matters.
79% of UK charities now use AI in daily work, according to the Charity Digital Skills Report 2026.
Now sit that next to a different number from the same report: 28% of charities say they don't know how to rate their board's AI skills at all. Not poor. Not good. Unknown.
The report says a third of boards are rated poor at AI skills, even as 79% of charities use AI every day. What it misses is what happens when you look at the trend, not just the snapshot.
The part the report states but doesn't connect
Two numbers from the governance section tell a different story. Only 19% of charities are reviewing data protection, security and safeguarding as they use more AI, down from 36% last year. Just 15% have updated their risk register for AI, down from 23% in 2025.
As AI use climbed to 79%, the number of charities checking safeguards nearly halved. The gap is widening, and it's happening in the same year AI use accelerated most.
This matters more than the skills numbers. If a charity is slow to build AI skills, and if safeguarding reviews fall as AI use rises, the organisation is moving backwards on the only thing that protects beneficiaries and donors.
What the skills numbers add to that picture
33% of boards and 25% of CEOs are rated poor at AI skills. Only 12% of charities say their board has strong AI skills. But the 28% who can't rate their board at all is the bigger issue. A poor rating means someone knows training is needed. Not knowing means nobody is looking closely enough to judge.
46% of charities, and 49% of small charities, have no trustee or board member with digital skills. Not weak skills—none.
On policy, 51% of charities have no AI policy. That rises to 62% for small charities. 44% are taking no action to manage AI risk or move forward with it.
Combine the skills gap with falling safeguards, and most boards can't assess AI risk. The checks are happening less often than last year, not more.
What "skilled" actually needs to mean here
You don't need a trustee who can prompt engineer or explain how a model works, just like you don't need a trustee who can do the charity's books themselves.
Boards need enough AI literacy to govern effectively and ask the right safeguarding questions. They should know the risks, whether an AI policy exists, and if safeguarding checks are up to date as AI use grows.
Boards already meet this standard in finance. Trustees don’t need to be tech experts, just able to ask the right questions. Most boards haven’t reached this with AI because the right questions aren’t being asked and key checks are slipping.
Why this is a governance failure, not a training gap
Staff are adopting AI, but board awareness is lagging, and nobody is connecting the two. Most AI use is in everyday tasks: 60% of charities use it for summarising notes and writing emails, 57% for documents and reports, 47% for research. These are things staff do without sign-off, procurement, IT approval or mention in a board pack.
Outside the charity sector, this is called shadow AI: staff using AI tools without leadership knowing or approving. Charities face a sharper risk than most businesses. Trustees, unlike line managers, are personally responsible for the organisation's decisions and legal compliance, even if they weren't in the room when a staff member made a decision by pasting something into a chatbot.
The trustees remain responsible for decisions; AI doesn’t add new duties, but it does create more ways to fall short of the existing duties of care and compliance, especially as oversight and checks decline.
How trustees can close the AI governance gap
Trustees need someone to do three things that most charities aren't doing right now:
Find out what staff are actually using, not just the official tools. Day-to-day use is often different from what's on paper.
Write down what's acceptable, especially for beneficiary and donor data, and make sure the board sees and approves it—not just a staff member.
Keep the safeguarding review and risk register up to date as AI use changes, instead of letting them fall behind as the sector's numbers already show.
This is the same strategic ownership gap seen throughout the report. It's why only 4% of charities use AI strategically, even though 79% use it in some way.
As a fractional head of digital for charities, I work between the board and the staff using these tools, part-time or interim, without the cost of a full-time senior hire. For AI governance, that means:
Auditing what staff are actually using across the organisation, not just what's officially sanctioned, so trustees work from reality, not guesswork.
Writing an AI policy that's specific to your charity, not a generic template. This covers beneficiary data, grant writing, safeguarding and where AI shouldn't be used.
Keeping the risk register and safeguarding reviews up to date as AI use changes, so they don't quietly fall behind like the sector average.
Setting up a reporting schedule so AI use reaches the board regularly, just like finance or safeguarding, instead of only surfacing when something goes wrong.
Translating between trustees and staff, so the board can ask the right questions without needing to become AI experts.
The value isn't in adding more process. It's giving trustees a clear, documented answer to what the charity is doing with AI and who's watching it, ready before a funder, auditor, or the Charity Commission asks. If you want to see what board-visible AI use looks like in practice, that's covered in The Agentic Charity. If you'd rather talk through where your charity stands, get in touch and I'll help you work out what's needed.
Common questions about charity AI governance
Does a charity board need to be skilled at using AI?
No, not in the sense of using AI tools well. What a board needs is enough literacy to govern AI use, not perform it, similar to how a board doesn't need to be accountants but is expected to read accounts and question an auditor. That means being able to ask the right questions about risk, know whether a policy exists, and check the safeguards behind it are being kept current.
Are charity trustees legally responsible for how staff use AI?
Yes. Trustees hold existing legal duties including a duty of care and a duty to act in the charity's best interests, and the Charity Commission's guidance is explicit that decision-making responsibility can't be delegated to AI. This applies whether or not the board was aware of how AI was actually being used.
Is charity AI governance getting better or worse?
Adoption is growing quickly, but the checks meant to manage the risk of that adoption are not keeping pace. The Charity Digital Skills Report 2026 found that the proportion of charities reviewing data protection and safeguarding for AI fell from 36% to 19% year on year and risk register updates fell from 23% to 15%, even as AI use climbed toward 79%.
What is shadow AI in a charity context?
Shadow AI is AI use that happens without an organisation's formal knowledge or approval, typically individual staff using free tools like ChatGPT for daily tasks. 79% of charities are using AI in some way, while 28% of charities say they can't even rate how skilled their own board is at it, a sign that oversight isn't keeping pace with use.