Do not put sensitive beneficiary, donor or staff data into public ChatGPT. A charity can use AI on sensitive data only when the tool, the place the data is processed, and the person who checks the output are agreed first. That is usually a private system the charity controls, not a public chat tab.

  • No case notes in a public tab.
  • Know where the data goes.
  • A person still checks the output.
  • Trustees can explain the choice.

Sensitive data here means information about people the charity supports, donors, volunteers or staff, including health, care, safeguarding and other special category information. The ICO guidance on special category data requires a lawful basis and a separate condition before that information is used. The ICO guidance on AI and data protection expects the same care when an AI tool is the one doing the reading.

A public ChatGPT tab does not give the board that control. Staff should not paste case notes, care records or donor details into it to save time. The Charity Commission has also published a note on charities and artificial intelligence: it is a prompt for trustees to understand the tool they are allowing, not a permission slip for an open chat box.

Before any sensitive record meets an AI tool

Name the record. Say which information would be used, and whether any of it is special category data.

Name the place. Write down the system that will hold the prompts and the answers. If the answer is “a public website”, stop.

Name the person. Someone at the charity reads the output before it is relied on, sent, or filed.

Name the purpose. One job, such as a first draft from an approved extract. Not “see what it says about this family”.

Tell the board. Trustees should be able to explain the choice in ordinary language. The essential trustee guidance (CC3) keeps that duty with them.

Where the work can sit

A private system the charity controls, with approved sources and a human review before anything goes out, is the right shape for this kind of work. That is the idea behind private AI for charity data, and behind tools such as Ask Your Data, which answer from the charity’s own records with a log of what was asked.

If you are choosing a first project, prefer a job that can run on approved, limited sources. The trustee questions in 12 questions before an AI project are a practical board checklist. If you cannot explain the tool to the people whose information it would read, do not use it on those records.

If staff are already using a public chat

Say so plainly, and stop the practice for sensitive records. Then agree an approved route for the job people were trying to get done. A quiet habit is harder to govern than a named tool with a named reviewer. This is also one of the everyday trust failures described in can your AI be trusted.

Questions

Can we paste a case note into public ChatGPT?

No. A public chat is someone else’s system. Do not paste beneficiary, donor or staff records into it.

Can a charity use AI on sensitive data at all?

Yes, when the tool, the place the data is processed, and the person who checks the output are agreed first. Special category data needs a lawful basis and a separate condition under UK GDPR.

Where should sensitive work run?

On a system the charity controls, with approved sources and a human review before anything goes out.

If staff are already using a public chat on sensitive records, or you want a private place for one agreed job, book a free 40-minute call by phone or Teams.

Book a free call