AI and MCP Acceptable Use Policy Template

An AI and MCP acceptable use policy tells staff which AI tools they may use, what data they may share with them, which MCP connections are allowed, and which AI actions need a person's approval. The template below covers all of that in 14 short sections. Copy it, replace the bracketed items, and have your legal adviser review it before you adopt it.

This template is not legal advice. It is a practical starting point written by a web and AI agency. Laws on privacy, employment monitoring and records differ by state, country and industry. A qualified lawyer should review your final policy.

Most AI policies written in 2023 and 2024 cover chatbots: what you may paste in. MCP adds a new question: what the AI may do in your other systems. This template covers both. It was checked against the sources linked below on 28 October 2026.

Why MCP belongs in your AI policy

MCP lets AI apps connect to email, files, CRMs and accounting systems through MCP servers. Those servers act on a person's behalf. A policy that only says "don't paste customer data into ChatGPT" misses that.

OWASP's entry on shadow MCP servers describes "unapproved or unsupervised deployments" that bypass security controls. Among its mitigations, it recommends adding MCP requirements to AI use policies and requiring security sign-off before deployment.

The MCP tools specification says there "SHOULD always be a human in the loop with the ability to deny tool invocations." The approval rules in section 8 put that into policy language.

The template

Copy everything in the box. Replace each [bracketed] item. Delete what does not apply.

[COMPANY NAME] AI AND MCP ACCEPTABLE USE POLICY
Version: [1.0]     Effective: [date]     Next review: [date]
Policy owner: [name, role]     Approved by: [name, role]

1. PURPOSE
This policy explains how staff may use artificial intelligence (AI)
tools, including AI tools connected to company systems through the
Model Context Protocol (MCP), safely and responsibly.

2. SCOPE
This policy applies to all employees, contractors and temporary staff
who use AI tools for company work, on company or personal devices.

3. DEFINITIONS
AI tool: any app that generates text, code, images or actions using
  an AI model (for example, chat assistants and coding assistants).
MCP server: a connector that lets an AI tool read data from, or take
  actions in, another system.
Approved list: the list of AI tools and MCP servers approved for use,
  kept at [link].
Sensitive data: [customer personal data, payment card data, health
  information, employee records, passwords and keys, non-public
  financials, other categories].

4. APPROVED TOOLS
4.1 Use only AI tools on the approved list for company work.
4.2 Use company accounts, not personal accounts, for approved tools.
4.3 To request a new tool, contact [policy owner]. Do not use it
    until it is approved.

5. MCP CONNECTIONS
5.1 Connect only MCP servers on the approved list.
5.2 Do not install MCP servers on any device, or connect new ones to
    company accounts, without written approval from [policy owner].
5.3 New connections start with read-only access unless [policy owner]
    approves more.
5.4 Do not approve an install command you have not read in full.
5.5 Do not change approved server settings, permissions or versions.

6. DATA RULES
6.1 Do not enter sensitive data into AI tools unless the tool is
    approved for that data category on the approved list.
6.2 Never enter passwords, API keys, access tokens or payment card
    numbers into an AI tool.
6.3 Treat AI output as a draft. Check facts, figures and code before
    using or sending them.

7. CREDENTIALS
7.1 Connect MCP servers with your own company login, not shared
    accounts or shared keys.
7.2 Do not store keys or tokens in shared documents, chats or code
    repositories.
7.3 Report any key or token you believe was exposed to [contact]
    immediately.

8. ACTIONS THAT NEED HUMAN APPROVAL
An AI tool must not complete these actions without a person reviewing
and approving each one:
  a. Sending email or messages outside the company
  b. Making, approving or changing payments
  c. Deleting records, files or accounts
  d. Sharing files or data outside the company
  e. Changing permissions, settings or user access
  f. [Other actions specific to your business]
Do not turn on "always allow" for these actions.

9. PROHIBITED USES
Do not use AI tools to:
  a. Break the law or company policy
  b. Make final decisions about hiring, pay, credit or similar
     matters without human review [adjust with legal advice]
  c. Impersonate a person or create misleading content
  d. Bypass security controls, approvals or monitoring
  e. Process data in ways customers or staff have not been told about

10. MONITORING AND LOGGING
The company logs the use of AI tools and MCP connections, including
which tools were called and when, to protect company and customer
data. [Describe what is logged and who can see it, as your legal
adviser recommends.]

11. REPORTING PROBLEMS
If an AI tool does something you did not ask for, asks to install
something unexpected, or seems to follow instructions from an email,
document or web page, stop using it and report it to [contact] right
away. Reports made in good faith will not lead to discipline.

12. TRAINING
Staff who use AI tools must complete [training name] before access
and [annually].

13. EXCEPTIONS
Exceptions must be approved in writing by [policy owner], recorded,
and reviewed at least every [6 months].

14. ENFORCEMENT AND REVIEW
Breaches may lead to removal of access and [disciplinary action in
line with company policy]. [Policy owner] reviews this policy at least
[annually] and after any significant AI incident.

Acknowledged by: ____________________   Date: ____________

How to fill in the key sections

Section 3: sensitive data

List your real data categories. Use the ones from your privacy notice and contracts. A short, specific list works better than "confidential information."

Sections 4 and 5: the approved lists

Keep the lists outside the policy, in a sheet or wiki page, so you can update them without re-issuing the policy. For each MCP server, record the publisher, version, access level and owner. Our MCP security checklist and server security review explain what to check before approving one.

Section 8: approvals

This is the most important MCP section. Tailor the list to your systems. If your team grows tired of approving, look at why. Our guide to approval fatigue covers ways to keep approvals meaningful.

Section 10: monitoring

Be honest and specific. Rules on telling employees about monitoring vary by location, so this is a section for your lawyer. Our guide to MCP audit logging explains what is typically recorded.

Fitting the policy into your existing framework

You likely already have an information security policy, a privacy policy and a vendor review process. This policy should point to them, not repeat them.

If you use the NIST AI Risk Management Framework, this policy is one way to put its governance ideas into practice. NIST describes the framework as "intended for voluntary use" and has published a Generative AI Profile (NIST-AI-600-1). For the wider governance picture, read MCP for IT leaders.

Rolling it out

1. Fill in the template and the approved lists.
2. Have legal and HR review it.
3. Brief managers first, then all staff, with examples from your own tools.
4. Collect acknowledgments.
5. Review after 90 days, then on the schedule in section 14.

Next step

Need help turning this policy into real controls, such as approved lists, read-only defaults and logging? Our MCP readiness audit maps your current setup against it. More guides are in the MCP security center.

Frequently asked questions