Book a Complimentary Session

Responsible AI policy starter

A practical AI policy your team might actually read.

A short policy your team can make its own. Fill in the parts in square brackets, and use the notes as a guide.

Not legal advice. Check your finished policy against the laws, regulations and contracts that apply to you, and your industry's and organization's own requirements. Ask someone qualified where you're unsure.

The policy, in ten parts

  1. Why we have this policy

    At [Organization], we use AI to do our work well. This page sets out the tools we use, how we use them, who is responsible and where to go with questions.

    Note: a page your staff can finish over a coffee gets read. A long document gets filed.

  2. Who is responsible

    [Role] owns this policy and our list of AI tools. Each important AI system has its own owner, recorded in that list. Everyone who uses AI in their work follows this policy.

  3. Tools we use

    Staff use the AI tools on our list. Ask [role] if you would like a tool added.

    Note: before a tool goes on the list, read the vendor's terms on how it stores and uses what you type into it.

  4. Sensitive information

    Confidential, personal, [customer or client], employee and regulated information, and anything a contract restricts, goes only into tools on our list that are cleared for that kind of information. Passwords, authentication codes, API keys and other secrets do not go into AI tools.

    Note: add two or three everyday examples from your own work. People remember examples long after they forget definitions.

  5. Responsibility for the work

    Each of us is responsible for the work we do with AI. Important external communications and consequential decisions, such as a decision about a person, get appropriate review before they go out or take effect.

  6. Checking the facts

    AI can produce convincing information that is incomplete, outdated or wrong. Check important facts against an appropriate source rather than assuming confident-looking output is accurate.

  7. AI systems we rely on

    For each important AI system, we write down its purpose and owner, the data and access it has, what it may do, what it must not do, where people become involved, how we review its performance and problems, and how it can be paused.

    Note: if this is hard to write down for a system, the system needs more thought before it runs.

  8. Being open about AI

    We do not present AI as a person. When someone interacts directly with AI, or when disclosure is required by law, contract, policy or the nature of the use, we make that clear appropriately.

    Note: some laws and contracts ask for this. Check the ones that cover you.

  9. Raising a concern

    If AI gets something wrong, or you're worried about how it's being used, tell [role or email]. We want to hear it.

    Note: people speak up when they know it's welcome. Say so plainly.

  10. Keeping it current

    [Role] looks at this policy every [six months], and whenever we start using a new kind of AI tool. Last looked at: [date].

    Note: AI tools change fast. A recent date tells staff the policy still applies.

Bring us a business challenge.

Complimentary Business + AI Opportunity Session: 45 minutes with Brian Pinkus, Founder & CEO.