All articles

What your small business AI policy should include

Your staff are probably using AI already, whether you have rules or not. Here is how to write a short, clear policy that covers the eight things they need to know.

Key figures

Employees who uploaded sensitive company information to public AI tools
48%
Employees who used AI at work without knowing if allowed
56%
AI-using UK businesses with no AI policy
17%

Sources: University of Melbourne and KPMG, Trust, attitudes and use of AI, 2025 · UK Government, UK business data survey, 2026

A good policy fits on two pages. It names your approved tools and lists the information that must never go into them. It also explains which outputs a person must check and tells your team who to ask.

A 2025 survey of 48,340 people in 47 countries by the University of Melbourne and KPMG found that 48 to 49% of employees had uploaded sensitive company information into public AI tools. They also found 56% had used AI tools at work without knowing whether it was allowed.

The UK Government's 2026 business data survey found 17% of AI-using businesses had no AI policy in place. A policy turns guesswork into a clear answer.

Banning AI rarely works

A 2025 University of Melbourne and KPMG survey concluded outright bans on generative AI may be ineffective. Risky behaviours, like uploading sensitive information, occurred in 67% of employees whose organisation banned AI, compared with 33% where there was no policy. A policy works better when it gives people approved tools to use.

Risky AI use was highest where AI was banned
  • Organisation banned generative AI67%
  • No policy in place33%

Sources: University of Melbourne and KPMG, Trust, attitudes and use of AI, 2025

What to keep out of AI

The UK Government's AI Playbook tells civil servants that when using public AI applications, they must not enter official information unless it is published or cleared for publication. You can set the same rule for your own confidential information.

Unless a tool is approved for it, your policy should exclude customer personal data, financial records, passwords and access details, confidential contracts, and anything covered by a non-disclosure agreement.

What the policy covers

  1. 01State why the business uses AI in one clear sentence.
  2. 02Which tools are approved and who can approve new ones.
  3. 03Information that must never be put into AI tools.
  4. 04Tasks where a person must check and verify AI output.
  5. 05Decisions that AI must never make on its own.
  6. 06How to label work created with help from AI.
  7. 07How to report a mistake or a potential data leak.
  8. 08Who owns the policy and when it is reviewed.

Keep it short. Review regularly

A two-page policy people read beats a twenty-page one they do not. Name one owner, review it every six months or when tools change, and tell staff what changed. If it touches personal data, check it against ICO guidance.

Our free AI readiness scorecard asks if your business has written rules on using AI. Governance is one of five areas our AI Readiness Audit always looks at.

Sources

Clinton Onyekwere

Clinton Onyekwere founded Clinton AI in Leeds. We help organisations adopt AI and build lasting internal skills.

Get our newest articles

We only email you when a new piece goes live.

See where AI fits your business.