Generative AI is now part of everyday working life. Employees use it to draft emails, summarise documents and work more efficiently.

But there is a risk hiding in plain sight. A survey by Deloitte found that 63% of workers in the United Kingdom are using AI tools in the workplace, even where this is unsanctioned. Using unapproved tools, often through personal accounts, may expose confidential or customer information to systems the business has not assessed. This is often described as "shadow AI".

The issue is not that businesses should ban AI. Used responsibly, it can be a valuable tool. The challenge is to make sure employees know what they can use, what information they can share, and when they need approval.

The risk is in the prompt

The biggest risk may be what someone puts into an AI tool, not the answer it gives.

Picture a colleague pasting a customer email into a chatbot to draft a reply, uploading a contract for a summary, asking a tool to analyse a spreadsheet or using AI to improve a proposal. In doing so, they may disclose information your business is required to protect, including:
  • customer, supplier or employee personal information;
  • confidential commercial terms, pricing or business plans;
  • legal advice, dispute information, or internal investigations;
  • product roadmaps, source code, or trade secrets; and
  • credentials, security information, or financial data.

Once information enters an unapproved tool, you may not know where it is stored, who can access it or whether the provider may use it to improve its services. Personal accounts add to that uncertainty.

Good intentions are not a complete safeguard

Most employees do not mean to expose company information. They may not realise that brainstorming an idea is different from uploading a customer document.

A general confidentiality obligation may not answer practical questions: Can I use this tool or upload this document? Can I use a personal AI account for work or include AI-generated content in a client deliverable?

What should an internal AI policy cover?

Your internal AI policy does not need to be long or overly technical. It does need to reflect how your business works and give your team clear direction. Here are the essentials:

Approved tools and accounts. Identify permitted tools and require company-approved accounts. Unapproved tools should not be used with company information.

Prohibited information. Keep confidential, client, personal, financial, and proprietary information, and credentials, out of unapproved tools. Set clear boundaries for approved tools too.

Human review. Ask your team to check output for accuracy, confidentiality, bias, intellectual-property concerns, and suitability before using or sharing it.

Higher-risk activities. Set additional controls for recruitment, performance management, legal or regulatory analysis, customer advice, and credit decisions, where errors could have serious consequences.

Approval and reporting. Make it easy to request approval for new uses and report possible uploads of protected information to unauthorised tools, so you can respond promptly.

Training and accountability. Use short, practical training and connect the policy with your existing IT, security, confidentiality, disciplinary and data-protection processes.

The contract matters too

Before approving an AI tool, check the provider's terms and what happens to your information.

Can the provider retain or use your information to train its models? Where is it processed, and by whom? What security and incident-reporting commitments apply? Can you retrieve or delete it when the relationship ends?

It is worth checking the small print. Standard click-through terms may give the provider broad rights to use data while offering limited warranties and low liability caps. Those terms may not be appropriate if the tool will handle sensitive customer, employee, or commercial information.

Make responsible AI use the easy option

Shadow AI thrives without safe tools or clear guidance. Provide approved alternatives, set limits, and make it easy to ask questions.

AI does not need to be the enemy. Put clear responsibilities, appropriate controls and protective contracts in place.

If you are working out how to use AI safely, Ethiqs Legal can help you review AI-provider and technology contracts, assess data-use and confidentiality risks, and develop practical contractual protections. Book a free 20-minute call with the Ethiqs team.

Book a free 20min Chat

Feel free to ask for details, don't save any questions!

Our Office

Business Hours

  • Monday - Friday - 9am to 5pm

Get in Touch

Ethiqs is committed to providing our clients with accessible, transparent and affordable legal services and this starts all the way from the initial consultation.