Skip to content
Welcome
UK
Dragonfly ITDragonfly Insights
Cybersecurity

Shadow AI: Many of Your Staff May Be Using AI Tools You Never Approved

Several 2025 and 2026 surveys suggest a large share of UK workers use AI tools their employer has not approved. Banning AI rarely works. Giving staff an approved tool with data controls usually does.

5 min readDragonfly IT

Explainer. Written by Dragonfly IT. Drafted with AI help and checked against the sources listed on this page.

Shadow AI is the use of AI tools at work that the business has not approved or does not know about. It is usually well meant: someone wants to write an email faster or summarise a long document. The risk is that company or client information ends up in a consumer tool with no business data protection, and nobody knows it happened.

The figures vary by survey and some come from vendors, but they all point the same way: unapproved AI use at work is common in the UK.

What do the UK surveys say?

Four pieces of research give a sense of scale:

  • Deloitte's 2026 UK workforce research, with fieldwork in May and June 2026, found that 17% of UK generative AI users, about one in six, pay for their own tools, with UK workers spending an estimated £1 billion a year on tools they use for work. It also found 31% of users use the technology without their employer knowing.
  • Research by Censuswide for Microsoft, among 2,003 UK employees in October 2025, found 71% had used unapproved consumer AI tools at work and 51% did so weekly.
  • A Turbotic study of more than 1,000 UK senior decision makers, reported by Consultancy.uk in September 2026, found 51% fear staff are entering sensitive company data into unauthorised AI tools.
  • The government's Cyber Security Breaches Survey 2026 found that only about a quarter (24%) of businesses using or considering AI had cyber security practices to manage AI risks.

Why banning AI rarely works

If staff find a tool useful, a ban tends to push its use onto personal phones and home laptops, where the business has even less visibility. The safer route is to give people an approved tool that does what they need, with the business's data protections switched on, and to be clear about what must never be pasted into anything else.

Whichever tool you choose, the principle is the same: approved, signed in with a work account rather than a personal one, set up with the business's data protections, and covered by a short written rule.

A one page AI policy outline

A short policy is better than a long one nobody reads. It can cover:

  • Which AI tools are approved, and that staff sign in to them with their work account.
  • What must never go into any AI tool: personal data about clients or staff, passwords, financial details and anything marked confidential, unless the tool is approved for it.
  • That AI output is checked by a person before it is sent, published or relied on.
  • Who to ask before trying a new AI tool or browser extension.
  • That staff will not be in trouble for asking, or for admitting they have used something unapproved.

What to do this week

Small steps that take an afternoon:

  • Ask staff, without blame, which AI tools they use and what for. The answers show what an approved tool needs to do.
  • Choose an approved tool and make sure staff sign in with work accounts.
  • Write and share the one page policy above.
  • Apply sensitivity labels to your most confidential documents if you use Microsoft 365, so their protection travels with them.
  • Review which browser extensions are installed on work devices, as some add AI features that send page content elsewhere.

Where Dragonfly IT can help

The earlier Insights articles on an AI policy for staff and on shadow IT go further on policy and discovery. For help choosing and setting up an approved AI tool with data protection, see the Dragonfly IT AI and digital transformation service, and Copilot licences are available through the Licensing range at Darter Tech.

Frequently asked questions

What is shadow AI?

Shadow AI is the use of AI tools at work that the business has not approved or does not know about, such as a personal chatbot account used to draft client emails. The main risk is that company or personal data goes into a tool without business data protection, outside the organisation's view and control.

How common is unapproved AI use in the UK?

Surveys vary, but they agree it is common. Research by Censuswide for Microsoft in October 2025 found 71% of 2,003 UK employees had used unapproved consumer AI tools at work. Deloitte's 2026 research found 31% of UK generative AI users use it without their employer knowing.

Should a small business ban AI tools?

A ban on its own tends to push use onto personal devices, where the business has less visibility. A better approach is to approve a tool that meets staff needs, require work accounts, set clear rules on what data must never be entered, and make it easy for people to ask before trying something new.

What should never be put into an unapproved AI tool?

Personal data about clients or staff, passwords and access details, bank and payment information, and anything marked confidential or commercially sensitive. If in doubt, staff should ask first. Approved tools signed in with a work account can be configured with the business's own data protections.

Sources

More in Cybersecurity

Is your business exposed to IT risks you can't see?

Our engineers audit your network, cyber security, backups and Microsoft 365 setup, then hand you a written report on where you're exposed and what we'd prioritise first.

Book My Free Risk Assessment

No cost · No commitment · 0333 0444 323