Privacy & Control
Shadow AI in UK businesses: what it is and how to manage it
Shadow AI in UK businesses is staff using AI tools nobody approved. 71% of UK employees have done it. Why bans backfire, and a one-page policy to use.
Last verified . AI memory features change quickly; dated claims were checked on this date.
Shadow AI is staff using AI tools their employer never approved. In the UK it is the norm: 71% of employees have used unapproved AI tools at work, according to Microsoft UK research with Censuswide published on 13 October 2025.
How common is shadow AI in the UK?
The study surveyed 2,003 UK employees. About half use unapproved tools every week. Microsoft sells enterprise AI tools, so weigh the source. The pattern matches what most owners already suspect.
Why shadow AI happens
The reasons are ordinary. The tools are familiar from home, and 28% of employees in the same study said their employer offers no approved option at all.
Why banning shadow AI backfires
A ban without an alternative pushes use out of sight. People keep using the tool that saves them time. They just stop telling you.
A one-page AI policy for small businesses
- Approved tools. Name them. Work happens in accounts the business controls.
- Never put in: client details, staff records, commercial figures, passwords.
- Check before it goes out. A person reviews anything sent to a client.
- Where knowledge lives. Decisions and client context are kept somewhere shared, not in personal chat histories.
- Who to ask. One named person for questions.
Give people an approved tool, so the easy option is also the safe one.
For what happens to that work when someone leaves, read what happens to an employee’s ChatGPT history when they leave. For what each assistant keeps about you, see AI memory.
Questions people ask
What is shadow AI?
Shadow AI is staff using AI tools their employer never approved. Microsoft UK research found 71% of UK employees have used unapproved AI tools at work.
Should a small business ban ChatGPT?
A ban without an alternative pushes use out of sight. People keep using the tool that saves them time and stop telling you. Give people an approved tool instead.
What goes in a one-page AI policy?
Approved tools, what never goes in, a check before anything reaches a client, where knowledge lives, and one named person to ask.
Sources
More from the Journal
Privacy & Control
What happens to an employee's ChatGPT history when they leave?
When an employee leaves, their personal ChatGPT history leaves with them. How to keep business knowledge when staff go, with a five-step checklist.
6 November 2026 · 3-minute read
Practical
Questions to ask your grandparents, in their own language
Ask in the language they think in. How to get family questions into simple spoken Punjabi, Urdu, Polish or another language, and keep the words they use.
9 November 2026 · 3-minute read
Memory
Your year in AI: what your ChatGPT and Claude history says you cared about
Your AI chat history is a rough diary of the year. How to read it privately from your ChatGPT or Claude export, and keep the memories that matter.
9 November 2026 · 3-minute read

