Most conversations about AI risk are about the future: job displacement, model bias, regulation still being drafted. Meanwhile there is a live exposure sitting in almost every organisation right now, and it has nothing to do with any of that.
Your staff are already using AI. Not through a procurement process, not with a policy, and not because anyone told them to. They are using free tools because those tools save them time and their workload has not reduced.
This is usually called shadow AI, and it is the most common AI governance gap in smaller organisations — precisely because nobody has decided anything, so nobody feels responsible.
What it actually looks like
Nothing dramatic. It looks like:
- A fundraiser pasting a donor list into a chatbot to draft segmented messaging
- An account manager summarising a client’s confidential file note
- Someone in HR asking a free tool to rewrite a disciplinary letter, with names in it
- A caseworker checking whether their write-up “sounds professional enough”
- A bookkeeper asking for help interpreting a spreadsheet of transactions
In every case, the person is trying to do their job better. That is worth saying clearly, because the instinct when this is discovered is to treat it as a discipline matter. It very rarely should be.
Why it matters legally
Client details, staff records, financial information, case notes or beneficiary data entering a public AI tool can constitute a personal data breach under UK GDPR. Whether it does depends on the tool, its settings and the terms attached to the account — and free consumer tools and paid business tools have materially different terms about what happens to what you type.
Most organisations have never checked which one their staff are using. That is the actual problem: not that people are using AI, but that nobody knows on what terms.
For a smaller organisation the stakes are disproportionate. UK GDPR penalties are calculated against turnover, and the reputational consequence of a breach involving beneficiary or client data is not something a small charity or a professional practice absorbs easily.
Sectors where this is more serious
If you handle sensitive personal data — health, children, finances, legal matters, safeguarding — this needs deciding before adoption rather than after. That includes care providers, clinics, advice services, schools, and any charity holding beneficiary records.
In those settings you may also need a data protection impact assessment before any AI tool touches the data, and you will certainly need a clear line on what must never go near a third-party tool under any circumstances.
The fix is genuinely small
Here is why this is worth doing this month rather than next year: it is the cheapest problem on this list to solve, and one of the most expensive to ignore.
- Ask, without blame. Find out what is being used and for what. You will get honest answers only if it is clearly not a disciplinary exercise. Frame it as “we want to make this safe, not stop it”.
- Write a one-page acceptable use policy. Not a twelve-page document nobody reads. One page: which tools are approved, what may never be entered, and who to ask.
- Publish a “never paste this” list. Concrete and specific to your organisation — names of beneficiaries, case notes, bank details, staff records, anything under legal privilege. People follow specific rules; they ignore general ones.
- Give people an approved option. This is the step most organisations skip, and it is the one that determines whether the policy works. If you ban the free tool without providing a sanctioned alternative, usage goes underground rather than stopping.
- Brief everyone for twenty minutes. Not an e-learning module. A short, human conversation about why it matters.
That is a week’s work at most, and for many organisations an afternoon.
The part that is easy to get wrong
A policy that bans AI outright looks decisive and tends to fail. Staff who were saving two hours a week do not simply stop; they stop telling you. You lose the productivity and the visibility, which is the worst of both outcomes.
The organisations that handle this well treat it as enabling something safely rather than prohibiting something risky. The tone of the conversation genuinely changes the result.
Where this sits alongside everything else
If you are considering AI more strategically — automating a process, improving how you handle enquiries, reducing time spent on reporting — this is the groundwork that comes first. Not because it is exciting, but because building on unmanaged usage means building on a compliance problem.
We include this in our free AI readiness assessment for exactly that reason. It would be difficult to advise anyone honestly about adopting AI while ignoring what is already happening in their organisation.
Questions we get asked
Should we ban staff from using AI tools?
Usually not. Staff who were saving two hours a week rarely stop — they stop telling you, so you lose the productivity and the visibility. Providing an approved option works better than prohibition.
How do we find out what AI tools staff are already using?
Ask, without blame. You will only get honest answers if it is clearly not a disciplinary exercise. Frame it as making current practice safe rather than stopping it.
Does using a free AI tool with client data breach UK GDPR?
It can. It depends on the tool, its settings and the data. Free consumer tools and paid business tools have materially different terms about what happens to what you type. Most organisations have never checked which they are using.