How to use ChatGPT at work safely in the UK

1. Start with your employer's policy

Many organisations now have an AI use policy, or rules within their IT or data protection policies. These may say which tools are approved (for example a company Copilot licence), which are not, and what kinds of data may be used with them. If there is no policy, ask your manager or IT team before using AI tools for work data. Using an unapproved tool for confidential work can be a disciplinary matter, even when the intention is good.

2. Protect personal data (UK GDPR)

Personal data means any information about an identifiable person: names, email addresses, phone numbers, customer records, HR details, health information and more. Under UK GDPR, your organisation must have a lawful basis and appropriate safeguards for how personal data is processed, including by third-party tools.

• Do not paste personal data into a public or personal AI account.
• Remove or replace identifying details before using text with an AI tool, unless your organisation has approved the tool for that data.
• Take extra care with special category data such as health, ethnicity, religion or trade union membership.
• If you are unsure whether something counts as personal data, treat it as if it does.
• The ICO (ico.org.uk) publishes guidance on AI and data protection if you want to read more.

3. Keep confidential information confidential

• Do not share client contracts, financial results, unreleased plans, passwords, security details or source code with tools that are not approved for it.
• Check the tool's settings. Some consumer tools may use your conversations to improve their models unless you opt out, while business versions often offer stronger data controls. Your IT team will know which applies.
• Summarise rather than paste. "A client in the retail sector is unhappy about a delayed delivery" is usually enough context for drafting help.

4. Check everything before you use it

AI tools can produce errors that sound authoritative: wrong figures, invented sources, outdated rules or misread instructions. You remain responsible for anything you send or publish.

• Verify facts, dates, names and figures against reliable sources.
• For UK rules (tax, employment, consumer law), check GOV.UK or a qualified professional, because rules change and AI tools may be out of date.
• Recalculate numbers yourself.
• Read the output as the recipient will, and remove anything you cannot stand behind.
• Never rely on AI alone for legal, medical, financial or safety-critical decisions.

5. Be careful with copyright and originality

Treat AI output as a draft that needs your judgement, especially for anything published or sold. Do not ask tools to reproduce copyrighted material such as articles, books or song lyrics, and do not present AI-generated work as someone else's. If your organisation has rules on AI-generated images or marketing content, follow them.

6. Be open about AI use

Transparency builds trust. Follow your organisation's approach to disclosing AI assistance, for example in reports, client work or recruitment. If a colleague or client would reasonably want to know that AI was involved, tell them. Never use AI to impersonate someone or to create misleading content.

7. Watch for bias and fairness

AI tools can reflect biases in the data they were trained on. Take particular care when AI is used in anything that affects people, such as recruitment, performance reviews or customer decisions. In those areas, a human should make the decision and be able to explain it.

A quick safety checklist

• Is this tool approved by my organisation for this kind of work?
• Have I removed personal and confidential information?
• Have I checked the facts, figures and sources?
• Would I be comfortable if my manager or client saw how I produced this?
• Is a human making the final decision?

Safe use is a skill employers value

Being able to say "I use AI to work faster, and here is how I keep it safe and accurate" is a strong message in a UK interview. If you want to test your practical prompting, including how you handle sensitive data in a scenario, the free PromptRank assessment is marked by a fixed rubric (no AI marking).