AI is already in the charities you support. Who is paying to govern it?

Ross McCulloch

Ross McCulloch

08 October 2026 | Estimated reading time: 4 minutes

naila-conita-m90HAx-fstU-unsplash

Most charities are now using AI. Very few have decided who is responsible for it. And most of the cost of that gap is being carried by the organisations least able to afford it.

That is the picture from the 2026 AI Equity Project, led by Meena Das of Namaste Data, with Beth Kanter as project advisor. It surveyed 880 nonprofits across the US and Canada. Around seven in ten report using AI, but only 12% have organisation-wide guidance and 21% have a formal process for responding if AI causes harm.

This is North American data from a self-selected sample, and the numbers shouldn’t be dropped into UK policy papers as if they were ours. Our legal context is different (UK GDPR, trustee duties under charity law), and so is our funding mix. But the underlying pattern is structural, not national. Anyone who works with UK charities will recognise it.

The report includes a section aimed specifically at funders and at sector networks and associations. Here’s what I think those findings mean for both in the UK.

For funders

Silence is not neutral. 43% of respondents said AI had not come up with their funders at all in the past year. Meanwhile, AI is being paid for through free accounts (36%), existing software (29%) and staff paying out of their own pockets (22%). Only 10% use funding specifically for AI. When funders say nothing, the risk doesn’t go away. It sits with the smallest organisations and, in some cases, with individual staff.

Fund the unglamorous parts. Asked what they would spend AI funding on, respondents put staff training first (49%), then governance (31%), dedicated staff or consultant support (28%) and change management (27%). Very few are asking for new tools. The preferred route was general operating grants (38%). This is the core funding argument again, applied to a new problem: governing AI well takes staff time, and staff time is what restricted project funding rarely covers.

Don’t make impact measurement the price of entry. Difficulty measuring AI impact rose from 35% to 57% in a year, and is now the most common barrier to securing AI funding. Funders who ask for a clear return on AI will favour larger organisations with evaluation capacity. A better question is the one Beth Kanter asks: if AI saves time, saved to do what? Funding charities to work that out is more useful than asking them to prove it upfront.

Be open about your own use. 19% of respondents said funders were using AI to review, score or process applications. Having spent five years on Foundation Scotland’s National Grants Committee, I know how much applicants read into every part of an assessment process. If AI is involved, applicants should know what it does, and there should be human review and a clear way to challenge a decision. For funders signed up to IVAR’s Open and Trusting commitments, this is a natural extension.

Watch the soft pressure. 18% had been offered funding that came with expectations about AI use, and 14% fear that not using AI will make them less fundable. Only 25% said their adoption had felt genuinely voluntary and well supported. Funders can help by saying plainly that not using AI will not count against an applicant.

Decide what you are responsible for. If you fund or recommend a tool and it causes harm, what do you owe the charity and the people it supports? Most funders haven’t answered that yet. It’s better to answer it before it comes up.

For sector infrastructure bodies

The UK has much denser support infrastructure than the US: national councils for voluntary organisations, local third sector interfaces and CVSs, and specialist digital organisations. That puts us in a better position to act on these findings, if we focus on the right things.

Move from templates to implementation. Around four in ten respondents have an AI policy, but only 12% describe AI use guided by organisation-wide practice. Those are different questions, so the comparison is indicative rather than precise. Still, it matches what I see: policies exist, but nobody owns them. The UK already has plenty of AI policy templates. What’s missing is practical help with ownership, approval routes and what to do when something goes wrong.

Treat boards as the gap. Board support is the third most important resource respondents say they need. Boards are involved in AI decisions in 27% of organisations, and 18% name board support as a barrier. Trustee-level understanding of technology and data is an infrastructure job. It can’t be left to individual charities to sort out board by board.

Help the sector learn from what goes wrong. There is currently no UK mechanism for charities to share AI harms, near misses or problems with suppliers. An anonymised, light-touch way of pooling those lessons would be new, and would save many organisations from repeating the same mistakes. The AI Equity Project itself keeps an open space for people to share experiences of data or AI harm, which is a useful model to learn from.

Offer independent spaces to test tools. Cost (43%) and compatibility with existing systems (39%) drive tool choice far more than reviewing a supplier’s privacy and security policies (26%). In practice, that means tools arrive by default through licences charities already hold. Infrastructure bodies are well placed to provide procurement questions and spaces to try tools without a sales pitch.

Build UK evidence. This report is useful, but it isn’t ours. Before using any of these figures in a policy ask or funding case, compare them with UK sources such as the Charity Digital Skills Report and Scottish sector data. Where we don’t have UK evidence, the gap itself is worth naming.

What I’d like to see

Only 7% of respondents said the communities they serve have meaningful input into AI decisions. The UK sector talks a great deal about lived experience and participation. Whether that extends to technology choices is a fair test of how seriously we mean it.

None of this needs a new national strategy. It needs funders to pay for the time governance takes and to be open about their own use. It needs infrastructure bodies to focus on ownership, boards and shared learning rather than more documents. And it needs all of us to stop treating AI as a technical question when it’s a question about responsibility. Technology choices are values choices, and that’s as true for the people funding and supporting charities as it is for the charities themselves.

If you are a funder or infrastructure body thinking about any of this, I’d be glad to compare notes.

The 2026 AI Equity Project report is by Meena Das (Namaste Data), with Beth Kanter as project advisor. Download the full report

More from our blog

Take your tech expertise into the boardroom

Digital Trustees and UKBT Institute are partnering to develop a new programme for professionals working across business, digital, data, design, AI and technology. The programme will help participants understand how charity boards operate, recognise the value their existing experience can bring at board level, and explore opportunities to become trustees or board advisers. Digital Trustees…

ScotlandIS and Third Sector Lab: strengthening digital leadership across Scotland’s third sector

We’re delighted to be working with ScotlandIS to help strengthen digital leadership across Scotland’s third sector, with an initial focus on our Digital Trustees programme. Digital is now fundamental to how charities and third sector organisations operate. From cyber security and data to AI, digital service delivery and organisational resilience, technology is increasingly shaping the…

What are the humans for?

Designing work when AI can do more of it By Ross McCulloch, Third Sector Lab I read a piece by Anna Aston recently, When Machines Do the Thinking, What is Left for Us?, and it has been rattling around my head ever since. It asks a question I think most organisations are avoiding, mostly because the…

How charities are using Microsoft Copilot to solve real problems

In 2024, the UK Department for Business and Trade gave 1,000 staff access to Microsoft 365 Copilot and ran one of the more substantial independent evaluations of the tool. The headline result looked promising: 72% of respondents were satisfied or very satisfied, and most reported some time savings. But the evaluation found no evidence that…

How Charities Are Using Claude to Solve Real Problems

A 2026 study of 346 nonprofits by Virtuous and Fundraising.AI found that 92% had adopted AI in some form, but only 7% said it had meaningfully expanded what their teams could accomplish. Closer to home, the 2026 Charity Digital Skills Report – based on 807 UK organisations – found much the same shape: 79% of…

Digital Trustees – September 2026 Cohort – Fully-funded places for The National Lottery Community Fund grant-holders.

Secure a fully funded place on the Digital Trustees June 2026 Cohort and get practical support to strengthen your board with digital, data or design expertise. Across two focused sessions, up to 30 organisations will work with the Third Sector Lab team to understand what digital skills their board needs, build confidence around digital governance,…