St. Louis · Kansas City

AI Acceptable Use Policy: What to Put in It and a Plain-English Starting Point

Your team is almost certainly using AI at work already. Someone drafts emails with a chatbot, someone else lets a free note-taker join client calls, and nobody has written down what is fine and what is off limits. You need a policy, probably this week, and you would rather not wade through a 20-page legal template to get one.
An AI acceptable use policy is a short written set of rules that tells employees which AI tools they may use for work, what information they must never put into them, and who to ask when they are unsure. For most small and mid-sized businesses it fits on one to three pages. It works best alongside approved tools, basic training, and a named person who keeps it current.
In this article
- What an AI acceptable use policy is
- Why write one now
- What a workable AI policy covers
- A plain-English starting policy
- Free tools versus work accounts
- What a 2026 employee AI policy should name
- How long an AI policy should be
- How to explain the policy so staff follow it
- How to keep your AI policy current
- Do you legally need an AI policy?
- How a managed service can help
- Frequently Asked Questions
What an AI acceptable use policy is
An AI acceptable use policy is the document that sets the everyday rules for using artificial intelligence at work. It covers chatbots, writing assistants, meeting recorders, and the AI features now built into software your team already uses.
It is a staff-facing document. It tells people what to do on a normal Tuesday, in words they can follow without a lawyer. The bigger questions, such as how the business chooses AI vendors or audits how AI is used, belong to AI governance, which the policy is one part of.
A policy is a statement of intent, and a written rule cannot stop a determined person from pasting a file into a website. That is why the policy works best paired with three other things: approved tools that make the safe choice the easy one, short training, and technical limits from your cybersecurity tools where your systems allow them. The policy is still the first and cheapest control you can put in place, and every other control points back to it.
Why write one now
Most employees are already using AI, and most have never been told the rules. In an EisnerAmper survey of 1,017 U.S. desk workers who use AI at work, published in August 2025, only 36% said their company had a formal AI policy. Just 41% said they tell their manager or ask permission before using AI.
That leaves a lot of people making their own calls. Industry surveys also suggest that a large share of employees who use AI at work rely on tools their employer never provided. Some of those outside tools are perfectly safe for some tasks. Without a policy, nobody knows which ones.
The risk usually comes from ordinary work. In a widely reported 2023 case, Samsung engineers put proprietary source code, chip test material, and a meeting recording into ChatGPT while debugging and summarizing. They were trying to get their jobs done faster, which is what your team is doing too.
What a workable AI policy covers
A workable policy answers nine questions, each in a sentence or two. Microsoft’s guidance on writing an AI policy employees can follow recommends a similar set: approved tools with a request process, acceptable and prohibited uses, data privacy and security rules, human review, disclosure, training, and a clear escalation path. SHRM adds bans on putting trade secrets and confidential information into prompts, and human review before AI work is published or acted on.
Here are the nine parts in plain terms:
- Scope. Who the policy applies to (employees, contractors, temps) and which tools count as AI.
- Approved tools. The AI tools your business has approved, on which accounts, and how to request a new AI tool.
- Never-enter data. The information that must never go into any AI tool that is not approved for it.
- Human review. A person checks AI output before it goes to a customer, gets published, or drives a decision.
- Disclosure. When staff must say that AI helped produce something, such as customer communications or published work.
- Reporting. What to do, and who to tell, if something sensitive was entered by mistake or the AI got something badly wrong.
- Owner. One named person who keeps the policy and answers questions.
- Consequences. What happens when the rules are broken, stated calmly.
- Review date. When the policy will be looked at again.
NIST’s generative AI profile backs up the reporting piece. It calls for acceptable use policies along with documented ways to track, respond to, and communicate AI incidents and errors. In a small business, that can be as simple as “tell the owner the same day.”
A plain-English starting policy
The sample below is a starting point you can copy and adjust. Replace everything in brackets, and keep the wording as plain as your team talks. This is general information, so have your attorney look it over if you handle regulated data.
[Company Name] AI Acceptable Use Policy
Effective date: [date]. Policy owner: [name, role, email]. Next review: [date].
1. Why we have this policy. AI tools can save time. They can also expose customer and company information if used carelessly. This policy explains how we use AI safely at [Company Name].
2. Who it covers. Everyone who works for or with [Company Name], including employees, contractors, and temporary staff, on any device used for company work.
3. What counts as AI. Chatbots, writing and coding assistants, image tools, meeting recorders and note-takers, browser extensions with AI features, and AI features inside software we already use.
4. Approved tools. You may use only the tools on our approved list, signed in with your company account: [Tool 1, account type]; [Tool 2, account type]. The full list lives at [location]. To request a new tool, email [owner] with the tool name and what you want to use it for. We aim to answer within [number] business days.
5. Never enter these into an AI tool unless the tool is approved for that type of information: customer or client names, contact details, or records; passwords, access codes, or security settings; health, financial, or payroll information; employee records; contracts, pricing, or anything under a confidentiality agreement; source code or internal system details; anything you would not post on our public website.
6. Check the output. AI makes mistakes and sometimes makes things up. You are responsible for anything you send, publish, or decide using AI. Check facts, figures, names, and links before using the result.
7. Say when AI helped. Tell your manager or the recipient when AI wrote a substantial part of work going to a customer, or when they would reasonably expect to know.
8. Meetings. Do not record or transcribe a meeting with AI unless everyone has been told and the tool is approved. Never record HR, legal, or security incident meetings.
9. Personal accounts. Do not use personal or free AI accounts for company work involving anything other than public information.
10. If something goes wrong. If you entered something you should not have, or AI output caused a problem, tell [owner] the same day. You will not be in trouble for reporting quickly. Hiding a problem is what makes it serious.
11. Not sure? Ask [owner] before you use it. A quick question is always welcome.
12. Breaking the rules. Misuse may lead to losing access to AI tools and to disciplinary action under our normal policies.
13. Acknowledgment. I have read this policy and will follow it. Name: ______ Signature: ______ Date: ______
That is roughly a page and a half once your details are filled in. It names an owner, lists the tools, spells out the data rules, and gives staff one place to go with a question.
Free tools versus work accounts
The single most practical rule for a small team is this: use business versions of AI tools, signed in with company accounts, for anything that is not public. Free or personal accounts should be limited to public, non-confidential tasks, or not used for work at all.
The reason is control. Consumer and business versions of the same tool often have different terms on whether your content is used to train models, how long it is kept, and what an administrator can see. Business tiers often add protections such as keeping company content out of model training, but the details differ by vendor and plan. Check each vendor’s current terms before you approve a tool, using these five questions to ask before you buy an AI tool as a guide, since those settings change.
There is a second reason that has nothing to do with training. When an employee uses a personal account, the business cannot see what was shared, cannot recover the history, and loses access when that person leaves. A work account keeps your information with your business.
What a 2026 employee AI policy should name
A current policy should name the tools people forget are AI. Chatbots get the attention, but the newer exposure comes from quieter places:
- Meeting note-takers and recorders. These can join calls automatically and keep full transcripts. Say when notice and consent are needed, and which meetings never get recorded.
- Browser extensions. Some AI extensions can read every page an employee opens. Treat extensions as tools that need approval, the same as any other AI tool.
- AI features inside approved software. Your email, document, and customer record platforms may have added AI features since you bought them. Decide whether those features are on, and for whom.
- Tools that connect to company systems. AI agents and assistants that link to your files, email, or business data can see far more than a chatbot. Approve these on what they can see, keep, and do.
That last point is the one to build the policy around. Brand names change often. Ask what a tool can see, what it keeps, and what it can change, and the rule still holds when the next tool shows up.
How long an AI policy should be
Keep it short. Most small-business guides suggest one to three pages, with a single page as a perfectly good start for very small teams. Microsoft’s guidance says AI policies are more useful when they are concise and practical.
Long policies tend to go unread. The trick is to keep the details out of the policy itself. The approved-tools list, setup instructions, and examples can live in a separate document that the owner updates without rewriting the policy every time a new tool is approved.
How to explain the policy so staff follow it
People follow a policy they understand and can use in ten seconds. Sending a PDF and asking for a signature is the start of a rollout, and the rest of the work is what makes it stick.
A rollout that works for a small team looks like this:
- Explain why first. Start with what could go wrong for a customer, in a real example, before any rules.
- Hand out a one-page quick guide. Include the approved-tools list and the “can I paste this?” test: public information is fine in approved tools, internal information goes only into approved work accounts, and customer data, passwords, health or financial records, and anything under contract stay out unless the tool is approved for that data.
- Train with scenarios. A 20-minute session working through real cases, such as summarizing a client email or cleaning up a spreadsheet, teaches more than a list of rules.
- Collect a signed acknowledgment. It confirms everyone has seen the policy and gives you a record.
- Make exceptions fast. If approving a new tool takes weeks, people will use it anyway. A quick yes or no keeps requests coming to you.
- Have managers go first. Staff watch what leaders do. A manager who asks before trying a new tool makes asking normal.
- Never punish someone for asking or reporting. The EisnerAmper figure is worth remembering here: fewer than half of employees say they ask before using AI. Make asking easy, and do not assume people will.
A 30-day check helps too. Ask the team which tools they wanted that were not on the list, and what was confusing. You will learn more in that conversation than in any audit.
How to keep your AI policy current
Review the policy on a set schedule and also whenever something changes. Microsoft’s guidance recommends regular reviews plus updates when new capabilities, security risks, regulations, or employee feedback come up.
Most small-business guides suggest a full review at least once a year, and some suggest every six months while AI tools are changing this quickly. Review it straight away after any of these:
- A new AI tool or a major new feature in a tool you already use.
- A change to a vendor’s terms on data use or retention.
- A new law or regulation that affects your industry.
- An incident, or a near miss someone reported.
Between reviews, a 15-minute look at the approved-tools list each quarter keeps it honest. Put the next review date in the policy itself so it does not drift.
Do you legally need an AI policy?
For most U.S. small businesses, no law requires a standalone AI policy. Existing laws still apply to how AI is used, though, including anti-discrimination, privacy, consumer protection, and health privacy rules such as HIPAA.
Some uses carry their own compliance duties. Using AI in hiring, lending, or housing decisions can bring specific obligations, and state rules are still changing, with Colorado among the states that have passed AI legislation. Businesses that serve customers in the European Union may fall under the EU AI Act. If any of that applies to you, talk with your attorney. Even where no law requires one, a written policy is the simplest way to show you took reasonable care.
How a managed service can help
If writing and rolling out a policy feels like one more project nobody has time for, a managed service can take most of it off your plate. You don’t need to have AI figured out. That’s our job.
At ThrottleNet, help writing an AI use policy and rolling it out to your team is part of ThrottleIQ, our Managed Intelligence service. We start from how your team actually works, including the tools they already use, and build the approved-tools list and data rules around that.
Our own AI is built to follow the same rules a good policy asks of staff. ThrottleIQ is read-only by default, which we call Zero-Write AI: it can read your connected systems and cannot change them, and that limit is enforced in how it connects rather than left to a setting. Your data stays inside your own Microsoft Azure account, every question asked is logged, and ThrottleNet says the AI model does not train on customer data. A policy still matters with tools like ours, since your team will use other AI tools too. It is easier to enforce when the approved tool is also the safe one.
If you are also wondering whether AI could take work off your team’s plate, our free written feasibility study is a separate way in. It follows a 30-minute discovery call, arrives within five business days, and is yours to keep. Start with a conversation, not a commitment.
Frequently Asked Questions
Who should own an AI acceptable use policy?
Usually the person who already looks after technology, operations, or cybersecurity, whether that is a staff member or an outside IT partner. The owner or top executive should sign off, HR should review the parts that affect employment, and legal should be consulted for regulated data or AI used in hiring decisions. Put the owner’s name and contact details in the policy, since a policy that says “ask IT” with no name gets fewer questions.
What is the difference between an AI acceptable use policy and AI governance?
The use policy is written for staff and covers what they may do day to day. AI governance is the wider system the leadership team runs: how AI vendors are chosen and approved, how use is monitored, how risks are assessed, and who is accountable. A small business can start with the policy alone and grow governance around it as AI use expands.
What should we do if someone already put customer data into a free AI tool?
Start by finding out what was entered, which tool and account were used, and when. Delete the conversation in the account if the tool allows it, check the tool’s terms on retention, and decide whether the customer or any regulator needs to be told under your existing privacy or contract obligations. Thank the person for reporting it, then use what happened to update the never-enter list and your training examples.

