A one-page AI policy for small businesses (plus what never to paste into ChatGPT)

The short answer

A small business AI policy should fit on one page and answer four questions: which AI tools staff may use, what information they may never put into them, which outputs a person must check before they reach a customer, and who to ask when unsure. Copy the template below, fill in the blanks, pair it with the red, yellow and green data list, and go through it with your team in a 15-minute meeting. You can do this today. It doesn't need a lawyer to get started, though regulated businesses should have theirs reviewed.

Why you need one before you need one

Your team is probably already using AI, on personal accounts, for work. That's not a disaster, but it means customer details, pricing and private documents may be going into tools with settings you've never checked. A one-page policy turns guesswork into a simple rule everyone knows. It also makes training easier, because you're teaching one approved way of working.

The one-page template

[Business name] AI use policy · Effective [date] · Owner: [name]

1. Approved tools. Staff may use [tool and plan, e.g. our company ChatGPT Business / Claude Team / Gemini account] for work. Do not use personal AI accounts for business information.

2. Never paste (red list). Customer payment details, Social Security or ID numbers, passwords and access codes, health information, employee records, and anything a client has told us is confidential.

3. Ask first (yellow list). Customer names with job details, contracts, financial reports, pricing and supplier terms.

4. Fine to use (green list). Our public information, general questions, drafts of marketing copy, templates with no personal details.

5. Human check. A person must read and approve anything AI drafts before it goes to a customer, is posted publicly, or is used for pricing, hiring, legal or money decisions.

6. Be honest. Don't present AI output as something it isn't. No AI-written reviews or testimonials, ever.

7. When unsure, ask [name]. Report mistakes quickly. That's how we improve this policy, and nobody gets in trouble for asking.

Review date: [every 6 months]

The red, yellow and green data list

Red: neverYellow: ask firstGreen: fine
Card and bank detailsCustomer names with job detailsYour website and public info
ID and Social Security numbersContracts and agreementsGeneral how-to questions
Passwords, alarm and access codesFinancial reportsMarketing drafts
Health and medical informationPricing and supplier termsTemplates with no personal data
Employee HR recordsInternal emailsRewriting your own notes

Adjust the columns for your business. A law firm moves almost all client material into red, unless it's in a tool reviewed for that use. A coffee shop may have very little in red beyond payment details and staff records.

Is ChatGPT safe for business use?

It depends on three things: which plan you use, how its settings are configured, and what you put into it. Business and team plans from the major vendors come with different data terms and admin controls than individual plans. Read the terms for the exact plan you buy. Settings such as whether chats are used for model training vary by plan, so check them. And the red list matters more than the brand: the safest information is information you never pasted. For a comparison of plans, see ChatGPT vs. Claude vs. Gemini.

Extra steps for regulated businesses

  • Health practices: if a tool will handle protected health information, HIPAA generally requires a business associate agreement with the vendor. HHS publishes sample BAA provisions. No BAA means no patient information.
  • Law firms: the ABA's Formal Opinion 512 covers competence, confidentiality, communication with clients and fees when lawyers use generative AI. Check your state bar's guidance too.
  • Financial, insurance and mortgage businesses: check your regulator's and carrier's rules on customer data before using any AI tool with client files.
  • Anyone wanting a fuller framework: NIST's free AI Risk Management Framework is the standard US reference, though most small businesses won't need more than this page.

Rolling it out in 15 minutes

  1. Fill in the template and the data list.
  2. Set up the approved tool on a business account if you can.
  3. Walk the team through it, with three examples from your own business.
  4. Pin it where people work, and add it to new-hire onboarding.
  5. Review it in six months, or sooner if you add a tool.

Your business's AI plan recommends tools with your data risks in mind, and flags where your industry needs extra care. That way the policy and the tools are decided together.

Your AI plan

Want this worked out for your business?

My AI Plan · for your business
$497

One fixed price. Voice interview, 9-slide written plan and a 30-minute expert review call.

Get your AI plan · $497

Full refund any time before your interview.

  • Your role, your team and where your week actually goes
  • The software you already pay for and which AI features it includes
  • The tasks most likely to return time quickly in your kind of business
  • Tools ranked by impact and effort, with monthly cost and setup time for each
  • A four-day quick-wins plan and a next-phase list
  • A 30-minute review call with an AI expert to agree your first steps
How My AI Plan works

One conversation. A written plan within 24 hours.

01

Interview

Right after checkout you get a private interview line. Call when it suits you for a 15 to 30 minute conversation about your business.

02

Analysis

Your answers are mapped to specific tools and scored by impact and effort, with monthly cost and setup time for each.

03

Report

A 9-slide written plan arrives by email within 24 hours of your interview.

04

Expert review call

A 30-minute call with an AI expert to walk through the plan and agree your first steps. Included for every client.

FAQ

One-page AI policy: questions

Does a small business need an AI policy?

If anyone on your team uses AI tools for work, yes. A one-page policy that lists approved tools, data that's off-limits and when a person must check output prevents most problems.

What should never be put into ChatGPT?

Payment details, ID numbers, passwords and access codes, health information, employee records and client-confidential material, unless you're using a plan and agreement approved for that kind of data.

Is ChatGPT HIPAA compliant?

Don't assume any AI tool is suitable for patient information. HIPAA generally requires a business associate agreement with a vendor that handles protected health information. Check whether the vendor offers one for the specific plan, and don't use patient data without it.

How long should an AI policy be?

One page is enough for most small businesses. Longer policies tend not to be read. Regulated businesses may need an additional section reviewed by their attorney or compliance advisor.

Half an hour of conversation.
Your business's AI plan in 24 hours.