[Organization Name] AI Acceptable Use Policy
- Effective date
- [Effective date]
- Policy owner
- Head of IT
- Applies to
- All employees, contractors and third parties
- Review cycle
- Every year
- Next review
- [Next review date]
- Version
- 1.0
Purpose
This policy sets out how personnel at [Organization Name] may use AI tools for work: which tools are approved, what information must never be entered into them, how they must be configured, and what to do when something goes wrong. It is not a ban on AI. It is how [Organization Name] uses AI with confidence.
1. Scope
This policy applies to all employees, contractors, and third parties ("personnel") who use generative AI tools, large language model interfaces, AI coding assistants, AI search tools, or AI content generation tools (collectively, "AI tools") in connection with their work for [Organization Name], whether on company-owned or personal devices.
It applies to all AI tools, including but not limited to ChatGPT, Claude, Gemini, Microsoft Copilot, GitHub Copilot, Cursor, Perplexity, Grok, and similar services, regardless of whether accessed through a browser, desktop application, API, or embedded in another product. That includes AI features built into other software, such as meeting transcription, writing assistants and the AI features of email, CRM and document tools.
In this policy, "company information" means any information [Organization Name] creates, receives or holds that is not public, including information about clients, customers and personnel, as classified in [Organization Name]'s information classification policy where one exists.
2. Approved tools
[Organization Name] has approved the following AI tools for use with company information, subject to the conditions in this policy:
| Tool | Approved plan or account | Required settings |
|---|---|---|
| ChatGPT | Approved plan or account: ChatGPT Enterprise, company workspace | Required settings: Sign in with the company workspace account. Business data is excluded from model training by default; keep it that way. Apps and connectors only with approval. |
| Microsoft 365 Copilot | Approved plan or account: Microsoft 365 Copilot license | Required settings: Signed in with a work (Microsoft Entra ID) account, so enterprise data protection applies. Never a personal Microsoft account. |
All other AI tools are unapproved for use with company information. Using an unapproved AI tool with company information is a violation of this policy.
To ask for a new tool, or for an exception to this policy, contact the policy owner. Exceptions are granted in writing only.
Work with AI tools happens in the approved accounts above. Personal AI accounts, free or paid, must not be used for [Organization Name] work, and company information must never be entered into them. Using AI tools for personal tasks on a company device is permitted, provided no company information is included in any prompt or attachment.
Approved tools may be used with public and internal information. Confidential information may be used only where [Organization Name]'s information classification allows it for that tool. The information listed in Section 3 is not permitted in any AI tool, approved or not, unless an exception is documented under Section 3.
Personnel are responsible for the work they produce with AI tools. Check facts, figures, citations and code before relying on AI output or sharing it, and do not use AI output as the only basis for a decision about a person, such as hiring or performance.
3. Prohibited data
The following must not be entered into any AI tool, including approved tools, whether by typing, pasting, uploading a file or connecting an integration, unless a specific exception is documented and approved in writing by the policy owner:
Regulated data
- Protected health information (PHI) as defined by HIPAA: patient names, diagnoses, medical record numbers, dates of service, and any other individually identifiable health information
- Personally identifiable information (PII) of customers, employees, or other individuals: names combined with government IDs, financial account numbers, biometrics, or other identifying information
- Payment card data: primary account numbers (PANs), CVVs, track data and PINs
Confidential business information
- Client and customer records, including customer contracts, pricing schedules and contract terms marked confidential
- Unpublished financial results, projections or forecasts
- Trade secrets, M&A targets, deal terms and due diligence materials
- Proprietary source code and internal tooling, including algorithms and model weights
- Attorney-client privileged communications
- Personnel records, compensation data and HR investigation materials
Authentication material
- Passwords, API keys, tokens, certificates, private keys, or any other credential, including credentials inside code, logs or configuration files
When in doubt whether information falls into a prohibited category, treat it as prohibited and ask the policy owner before proceeding.
4. Required settings and technical controls
Personnel using approved AI tools must keep the following settings in place:
- Training opt-out. Turn off any setting that lets the AI provider use your prompts to improve its models. Business plans usually exclude business data from training by default; confirm it for each tool with the policy owner.
- Chat history. Turn off saved conversation history for any session that involves confidential information. Where history cannot be turned off, do not enter confidential information.
- Memory and personalization. Turn off AI memory features that store information across sessions when using AI tools for work, and delete any sensitive content already stored in memory.
- Connected apps and integrations. Do not connect AI tools to company systems (email, CRM, code repositories, cloud storage) without prior review and approval by the policy owner.
The policy owner checks these settings when a tool is approved and at each review of this policy.
Technical controls
Where an approved tool supports single sign-on, personnel sign in through [Organization Name]'s identity provider, so access can be reviewed and removed when someone leaves.
Personnel are encouraged to use a data-masking tool, which detects sensitive values, such as names, contact details, account numbers and credentials, on the device and replaces them before a prompt or file is sent. Masking reduces risk; it does not make the information in Section 3 permitted. Use an on-device masking tool, such as Locke, to keep sensitive values out of prompts.
5. Incident response
If you believe you have sent prohibited data to an AI tool, whether by typing, pasting, uploading a file or through an integration, take these steps immediately:
- Stop the session. Do not continue the conversation or add more context.
- Delete the conversation. Use the tool's delete feature to remove it from your history. This may not immediately delete it from the provider's servers.
- Report it within 24 hours. Tell the policy owner at [email address or chat channel]. Include the AI tool used, the kind of data involved, the date and time, and your best estimate of how much was shared.
- Do not try to fix it alone. Leave contact with the AI provider to the policy owner.
Failure to report a suspected incident is treated the same as the underlying policy violation.
Violations of this policy are handled under [Organization Name]'s disciplinary process, from coaching to a written warning to termination, applied consistently. Prompt, voluntary reporting is taken into account. For contractors and other third parties, a violation may end their access or their engagement.
6. Training and acknowledgment
All personnel must complete AI acceptable use training before first using an AI tool for work, and every year after that. Training can be a walkthrough of this policy with the policy owner. It covers:
- This policy and the prohibited data categories
- How to recognize confidential information in your own work
- The required settings for approved tools
- How to report an incident
Personnel must acknowledge this policy in writing (electronic acknowledgment is acceptable) when they first receive it and every year after that. The policy owner keeps acknowledgment records for the duration of employment plus three years.
7. Review and updates
This policy is reviewed and updated at least every year, and also when any of the following occur:
- A new AI tool is approved or removed from the approved list
- A significant change occurs in an approved tool's privacy terms or data handling
- A regulatory change introduces new requirements applicable to AI tool use
- An incident occurs that reveals a gap in the policy
The approved tool list in Section 2 is a living list. The policy owner reviews it every quarter and decides requests for new tools within 30 days.
Policy owner: Head of IT. The policy owner maintains this policy and the approved tool list. Questions about this policy go to [email address or chat channel].
Acknowledgment
I have read [Organization Name]'s AI Acceptable Use Policy, I understand it, and I agree to follow it.
About this draft
Drafted with the free AI use policy generator at sonomos.ai/tools/ai-use-policy-generator. It is a starting point, not legal advice. Before you adopt it, fill in anything in [brackets], check it against your other policies, contracts and the laws that apply to you, and have it reviewed by counsel. Then delete this note.