AI acceptable use policy template.
Your staff already use AI tools. A short policy tells them which tools are approved and what data may go in. Pick the global version or the India version built around the DPDP Act, then copy it or download it and make it yours.
What the policy settles
The template, in two versions.
Choose a version below. Each is a plain staff policy of about a page and a half once you fill in the brackets. This is general information, not legal advice, so have your own lawyer review the final text.
Global version
For companies that do not hold personal data of people in India, or that want one plain policy for staff in several countries. It is about tools, data levels, checking output and reporting problems.
Fill in every bracket. Delete clauses that do not fit.
1. Purpose and scope
[Company] wants its people to use AI tools well. This policy says which tools you may use and what you may put into them, so that we protect our customers, our colleagues and our own information. It applies to all employees, contractors and interns, on any device, and covers AI features inside other software as well as stand-alone chat tools.
2. What counts as an AI tool
Any software that generates text, code, images, audio, video or decisions from the content you give it. This includes chat assistants, writing and translation tools, code assistants, meeting recorders and note-takers, browser extensions, and AI features switched on inside our existing software.
3. Approved tools
The approved tools are listed in the AI tool register held by [owner, for example the head of IT]. Each one is used on a work account through single sign-on. If a tool is not on the register, ask [owner] before you use it. We will reply within [number] working days.
4. No personal or free accounts for work
Do not put company, customer, candidate or colleague information into a personal or free AI account, on any device. We hold no contract for those accounts and cannot control what happens to the data.
5. What data may go in
Level 1, no personal or confidential data: any approved tool. Level 2, ordinary personal or confidential data such as names, emails, support tickets and contract drafts: approved tools only, and only what the task needs. Level 3, high-risk data such as passwords, API keys, payment card numbers, health records, children’s data and salary data: never into a general chat tool. Remove it or mask it first, or use a tool [owner] has set up for it.
6. Secrets and source code
Never paste credentials, API keys, private keys or tokens into any AI tool. Source code goes only into code assistants on the approved list, with the training setting recorded as off.
7. Customer and partner material
Check the contract before you put customer or partner material into an AI tool. If a contract limits AI use, that limit wins over this policy. If you are not sure, ask [owner].
8. Check the output
You are responsible for anything you send or publish, whether or not an AI tool wrote it. Check facts, figures, quotes and code before you use them. AI tools can state false things with confidence. Where the law or a customer contract requires us to say that content is AI-generated, we say so.
9. Decisions about people
Do not use an AI tool to make or recommend a final decision about hiring, pay, promotion, discipline, credit or access to a service without a named person reviewing it and taking the decision. Record who reviewed it.
10. Meeting recorders and note-takers
Only use an approved recorder. Tell everyone on the call before it starts, and do not record a meeting where someone objects. Store transcripts in the approved location, and delete them after [period].
11. AI in our own products
A new AI feature, a new model provider or a change of model needs a review by [owner] before release. The review updates the AI tool register, the data flow diagram and any customer-facing description of the feature.
12. Report problems the same day
If you think data went into the wrong tool, an AI tool leaked something, or an output caused harm, tell [contact] the same day. You will not be blamed for a prompt report. [Company] decides the next steps under its incident plan, including any notice we owe to customers or regulators.
13. Training, review and breaches of this policy
Everyone who uses AI tools at work completes [training] when they join and [frequency] after that. [Owner] reviews this policy at least every [period] and when we adopt a major new tool or the law changes. A breach of this policy is handled under [HR or disciplinary policy].
India version, built around the DPDP Act
For companies that handle personal data of people in India. It adds vendor contracts, purpose, children’s data, deletion and breach reporting, in line with the DPDP Act, 2023 and the DPDP Rules, 2025. Most of the Act’s duties start on 13 May 2027, by our count.
Fill in every bracket. Have your own lawyer review it against your sector rules.
1. Purpose and scope
[Company] wants its people to use AI tools well. This policy says which tools you may use and what you may put into them, so that we protect our customers, our colleagues and our own information. It applies to all employees, contractors and interns, on any device, and covers AI features inside other software as well as stand-alone chat tools.
2. What counts as an AI tool
Any software that generates text, code, images, audio, video or decisions from the content you give it. This includes chat assistants, writing and translation tools, code assistants, meeting recorders and note-takers, browser extensions, and AI features switched on inside our existing software.
3. Approved tools
The approved tools are listed in the AI tool register held by [owner, for example the head of IT]. Each one is used on a work account through single sign-on. If a tool is not on the register, ask [owner] before you use it. We will reply within [number] working days.
4. No personal or free accounts for work
Do not put company, customer, candidate or colleague information into a personal or free AI account, on any device. We hold no contract for those accounts and cannot control what happens to the data.
5. What data may go in
Level 1, no personal data: any approved tool. Level 2, ordinary personal data such as names, phone numbers, emails, support tickets, CVs and call transcripts: approved tools only, and only what the task needs. Level 3, high-risk personal data such as Aadhaar numbers, PANs, bank and card details, health information, children’s data and salary data: never into a general chat tool. Mask it first, or use a tool [owner] has set up for it.
6. Contracts and vendor settings
An AI tool may only process personal data if [Company] holds a contract with the vendor on a business plan, and [owner] has recorded the training setting, the retention period and where data is processed. The vendor acts for us, but we stay responsible for the personal data it handles.
7. Use data only for the purpose it was given
Use personal data in an AI tool only for the purpose the person was told about, or another purpose the law allows. If you want to use customer data for a new AI purpose, ask [privacy contact] first.
8. Children’s data
Do not put personal data of anyone under 18 into an AI tool unless [privacy contact] has approved the use in writing.
9. Staff and candidate data
CVs, appraisal notes, payroll files and disciplinary records may only go into AI tools that [HR head] has approved for that use, and only what the task needs.
10. Secrets and source code
Never paste credentials, API keys, private keys or tokens into any AI tool. Source code goes only into code assistants on the approved list, with the training setting recorded as off.
11. Check the output
You are responsible for anything you send or publish, whether or not an AI tool wrote it. Check facts, figures, quotes and code before you use them. Do not use AI output as the final word on a decision about a person. A named person reviews it and takes the decision.
12. Meeting recorders and note-takers
Only use an approved recorder. Tell everyone on the call before it starts, and do not record a meeting where someone objects. A recording of a person’s voice is personal data, so store transcripts in the approved location and delete them after [period].
13. Deletion and rights requests
If a customer, candidate or colleague asks for their data to be corrected or erased, tell [privacy contact] the same day, and say which AI tools may hold it. We will check those tools as part of the request. Keep chat history on the shortest retention setting that works.
14. Report incidents at once
If personal data went into the wrong tool or an AI tool leaked something, tell [contact] immediately. Our incident plan covers both the CERT-In report, due within 6 hours of noticing a listed incident, and the personal data breach duties under the DPDP Rules, including the 72-hour report to the Data Protection Board from 13 May 2027. You will not be blamed for a prompt report.
15. AI in our own products
A new AI feature, a new model provider or a change of model needs a review by [owner] before release. The review checks which personal data reaches the model, whether it is masked, and whether the vendor is covered by a contract. It updates the AI tool register and our list of processors.
16. Training, review and breaches of this policy
Everyone who uses AI tools at work completes [training] when they join and [frequency] after that. [Owner] reviews this policy at least every [period], when we adopt a major new tool, and when the law changes. A breach of this policy is handled under [HR or disciplinary policy].
Nothing is stored or sent. The file is made in your browser.
Three data levels staff can remember.
Both versions use the same three levels. They are our working scale, not a legal category. The DPDP Act, for example, does not create a separate class of sensitive personal data, so the third level is a list of data where a leak does the most harm or where sector rules add duties.
Level 1: no personal or confidential data
Public text, generic code, marketing copy with no names. Any approved tool.
Level 2: ordinary personal or confidential data
Names, emails, support tickets, CVs, call transcripts, draft contracts. Approved tools only, and only what the task needs.
Level 3: high-risk data
Credentials, payment card numbers, health records, children’s data, salary data, and in India Aadhaar numbers and PANs. Not in general chat tools. Mask first, or use a tool set up for it.
Keep a register behind the policy.
The policy points to an AI tool register, and the register does the real work. Without it, “approved tools” is an empty phrase. For how to fill it in on a work account, with training off and retention set, see our guide on using ChatGPT at work under the DPDP Act.
Tool and owner
The name of the tool and the person who answers for it.
Account and plan
Work account on a business plan, with single sign-on. No personal logins.
Training and retention
Whether the vendor trains on your content, how long history is kept, and who set it.
Data level allowed
The highest level from the table above that the tool may receive.
Where data is processed
Taken from the vendor’s terms and the contract, with the date you checked.
Contract and review date
The agreement or data processing terms you hold, and when you will look again.
What a policy does and does not do.
A policy sets the rule and gives people something to point to. It does not show what is in use. Before you roll it out, find out which tools your staff already use, and give them approved ones that work, so nobody needs a personal account.
ISO/IEC 42001
Annex A starts with policies related to AI. A written policy is one control, not the whole management system, and accredited certificates come only from an accredited certification body. Mono prepares you; we do not certify.
NIST AI RMF
NIST’s AI RMF 1.0 is voluntary, and a revision is in progress. Its Govern function covers policies and accountability, which is where a policy like this sits. There is no NIST certification.
EU AI Act
Article 4 of Regulation (EU) 2024/1689 gives providers and deployers an AI literacy duty for their staff, and the AI Omnibus softened that duty. The Omnibus moved the high-risk deadlines but the obligations remain. A staff policy does not meet a provider’s duties for a high-risk system.
DPDP Act, India
The Act applies to the personal data in prompts, files and transcripts. You stay responsible for processing a vendor does for you (Section 8(1)), you need a valid contract with it (Section 8(2)), and security safeguards cover it (Rule 6). Those duties start on 13 May 2027, by our count.
If you also record meetings with AI note-takers, add a consent step to the policy. Our guide to AI meeting recorders and consent covers it.
Where Mono fits in.
Mono does not certify you, audit you or give legal advice. We find the AI tools in use, build the controls behind the policy and collect the evidence. If a buyer is asking for an AI policy, our shadow AI audit produces one that matches the tools you really run. For companies in India, DPDP and AI tools covers the personal data side, and ISO 42001 readiness prepares you for an audit by an accredited certification body.
- Step 1
Discover
We find the AI tools your people use, from sign-in records, expenses, browser extensions and a short staff survey.
- Step 2
Map
We trace which data reaches each tool, and on what terms.
- Step 3
Replace
We set up approved tools with the right settings, close the risky ones and write the policy to match what is really in use.
- Step 4
Keep watch
We leave you a simple way to spot new AI tools as they appear.
Related services.
Shadow AI audit
Find the AI tools your team uses without approval, then replace them with approved ones.
See the auditDPDP and AI tools
Where personal data of people in India reaches AI tools, checked against the DPDP Act.
See DPDP and AI toolsAI security questionnaires
A reviewed answer library for the AI sections of buyer questionnaires.
See questionnaire helpAI policy FAQs.
What should an AI acceptable use policy cover?
At minimum: which AI tools staff may use, which accounts they must use them on, what data may go into each tool, who checks the output, what to do when something goes wrong, and who owns the policy. Keep it short enough that people read it. A policy that nobody reads does not change what staff do.
Should we ban ChatGPT and similar tools?
We usually advise against a blanket ban. People who are blocked tend to move to personal phones and free accounts, where you see nothing and hold no contract. A better route is a short list of approved tools on work accounts, clear limits on data, and a way for staff to ask for a new tool.
Which version should we use?
Use the India version if your company handles personal data of people in India, because it adds the DPDP points: contracts with AI vendors, deletion, children’s data and breach reporting. Use the global version if it does not. A company with staff and customers in several countries can start from the global text and add the India clauses.
Is a template enough, or do we need a lawyer?
A template is a starting point. It does not know your tools, your contracts or your sector. Fill in every bracket, then have your own lawyer review the result. Mono is not a law firm and this page is general information, not legal advice.
Does having this policy make us compliant with ISO 42001 or the EU AI Act?
No. ISO/IEC 42001 asks for policies related to AI as one group of controls inside a wider management system, and accredited certificates come only from an accredited certification body. The EU AI Act has its own duties for providers and deployers that a staff policy does not cover. Treat this policy as one piece of evidence, not a certificate.
How do we know staff follow the policy?
A policy states the rule. It does not find the tools people use without telling you. Single sign-on records, expense data, browser extension lists and a short staff survey show what is really in use, and a shadow AI audit does that work. Run the check once, then repeat it when you review the policy.
Find out which AI tools your team really uses.
Sources: Digital Personal Data Protection Act, 2023, Sections 8 and 9. Digital Personal Data Protection Rules, 2025, Rules 6 and 7 (G.S.R. 846(E), Gazette of India, 13 November 2025). CERT-In Directions, 28 April 2022. ISO/IEC 42001:2023, Annex A. NIST AI RMF 1.0 (26 January 2023). Regulation (EU) 2024/1689, Article 4, as amended by the AI Omnibus (Official Journal, 24 July 2026). Last reviewed 20 November 2026. See also our AI governance page.