AI tools can help people write, analyse, research and create. Used carelessly, they can also expose confidential information, introduce errors and create risks that are difficult to unwind. These practical principles are a starting point for safer use at work.
Seven principles for safer AI use
1. Protect sensitive information
Do not enter personal data, customer records, confidential business information, credentials, source code or commercially sensitive material into an AI tool unless your organisation has approved that tool and the specific use. Follow your organisation’s data classification and handling rules.
2. Use approved tools for work
Check which tools your organisation has approved and what accounts or configurations you should use. A personal account may have different data settings, contractual protections and administrative controls from an enterprise account.
3. Verify important outputs
AI systems can produce convincing but incorrect answers, fabricated references and outdated information. Check facts, calculations, citations and recommendations against reliable sources before relying on them or sharing them.
4. Keep human judgement in the loop
Do not delegate decisions with meaningful consequences to an AI tool without appropriate human review. Be especially careful with decisions affecting employment, customers, access to services, finances, safety or people’s rights.
5. Think about copyright and ownership
Do not assume that material generated by AI is original, accurate or free of third-party rights. Follow your organisation’s rules for source material, licensing, attribution and review before publishing or incorporating outputs into products.
6. Watch for bias and unfair outcomes
AI outputs can reflect gaps or biases in data and design. Check whether the result is appropriate for the people and context involved, and raise concerns when an output appears discriminatory, misleading or unsuitable.
7. Be transparent and accountable
Follow any disclosure, record-keeping and escalation requirements that apply to your work. Know who owns the process and where to report an unexpected result, data exposure or suspected misuse.
Before you use an AI tool at work, ask:
- Is this tool approved for this kind of work?
- Am I about to share information that should remain private or confidential?
- Could someone be affected by the output?
- How will I check whether the answer is correct?
- Do I need to tell someone that AI was used?
- Who is accountable for the final decision or result?
A note for employers
Employee guidance works best when it is supported by clear organisational rules. Define approved tools, data-handling expectations, review responsibilities and a simple route for asking questions. Training should reflect the roles people perform and the AI systems they use.
How this relates to the EU AI Act
Article 4 of the EU AI Act requires providers and deployers to take measures to support the AI literacy of relevant staff and other people operating or using AI systems on their behalf. AI literacy is only one part of responsible AI use: privacy, security, employment, consumer, intellectual property and sector-specific rules may also be relevant.
Read our practical introduction to the EU AI Act →
Important: This guide provides general educational information. It is not legal, security or data-protection advice. Follow your organisation’s policies and seek specialist advice where appropriate.
Get the free AI safety guide
Enter your email address below to access the Lucentreg field guide — a concise reference for safer AI use at work.
We’ll use your email address to provide access to this guide. It will not subscribe you to marketing updates. Read our Privacy Policy for details about how your information is handled.
