AI safe-use policy.

1. What this policy is for

[Company] wants everyone to use AI where it helps, safely. This policy sets out which tools you can use, what information can go into them, and who to ask. It covers everyone who works for [Company], including contractors, on any device used for work.

2. The tools you can use

Use the approved tools below, through your [Company] account. If a tool you want is not on the list, ask [named owner] before you use it for work. We would rather add a useful tool than have people use it quietly.

• [Tool, for example Microsoft Copilot]: [what it is approved for]
• [Tool]: [what it is approved for]
• [Tool]: [what it is approved for]

3. What information can go in

Think of information in three groups. If you are not sure which group something is in, treat it as the stricter one and ask.

• Never: passwords and access keys; payment card details; health or other special category data about a person; anything a client has told us is confidential and not for AI tools.
• Only with approval from [named owner]: personal data about customers or staff; client documents and data; unpublished financial results; contracts and pricing.
• Fine in an approved tool: public information; your own drafts and notes; internal material that is not personal or confidential.

4. Accounts

Use your [Company] account for work, never a personal one. Company accounts are set up so our information is not used to train the tool, and so we can see what is in use. Do not share accounts or logins.

5. A person checks the output

AI can be wrong and sound certain. Before anything AI has produced is sent to a client, published, used to make a decision about a person, or relied on for a number, the person sending it checks it and is responsible for it.

6. Clients, confidentiality and ownership

Our contracts and our clients' wishes come first. If a client has said their material must not go into AI tools, it does not. Do not put other people's copyrighted material into AI tools to reproduce it. Work created with AI for [Company] belongs to [Company] in the same way as any other work.

7. If something goes wrong

If you think something has gone into an AI tool that should not have, or an AI tool has done something unexpected, tell [named owner] straight away. Reporting it quickly matters more than how it happened, and nobody is in trouble for reporting.

8. Who to ask, and when this is reviewed

Questions go to [named owner], [email]. This policy is owned by [role] and reviewed every [six months], or sooner when we add a tool or the law changes. Last reviewed: [date].