Home › Legal & compliance › AI acceptable use policy and tool approval: how to do it
AI acceptable use policy and tool approval: how to do it
From official vendor pages; selected tasks also include dated logged-out free-plan trials. Report an error.
General information only, not legal advice. A qualified professional should make the final decision.
Before you paste anything, remove names, contact details, ID numbers and other personal details, or replace them with placeholders like [Client A].
Intermediate Risk: Medium ⚖️ Legal & compliance
Draft a workplace AI acceptable use policy and a simple process for approving new AI tools: which tools are allowed, what data can go into them, who approves new ones, and what people must check before using AI output. AI can produce a solid first draft; your legal, privacy and IT people must adapt it to your laws, contracts and risks.
Who it’s for: Small-business owners, IT and security leads, HR and compliance teams setting AI rules for staff.
Why medium risk: A policy that's wrong for your jurisdiction or contracts gives false comfort; it needs review by someone accountable.
ChatGPT has the highest score on our published rubric (94/100) of the 4 tools listed for this task. OpenAI's pricing page describes Business and Enterprise plans for organisations (Business from 2 users) and lists encryption in transit and at rest (chatgpt.com/pricing, checked Oct 23, 2026).
Step by step
- List the AI tools people already use (ask; most teams use some without approval).
- Classify your data simply: public, internal, confidential, personal/regulated.
- Ask AI for a draft policy with sections for approved tools, data rules, human review, transparency and prohibited uses.
- Ask for a tool approval checklist: vendor, business need, data types, training/retention settings, admin controls, contract terms.
- Have legal, privacy, IT and HR review and adapt both; add your contacts and approval time frame.
- Publish a short register of approved tools and review it at least yearly or when a vendor's terms change.
AI prompts to write an AI acceptable use policy
Replace the [brackets] with your details. See all prompt examples
Draft an AI acceptable use policy for a [size]-person [industry] company in [country/province/state]. Sections: purpose and scope, approved tools (refer to a register), data rules by class (public / internal / confidential / personal), human review of output, transparency with customers, prohibited uses, reporting incidents, review date. Plain English, under 2 pages. Mark anything that depends on local law as [CHECK WITH LEGAL].
Create a one-page AI tool approval checklist for staff requests. Include: requester and business need, data types involved, whether the vendor trains on inputs and how to turn it off, data retention, admin/SSO controls, where data is stored, contract terms, integrations, and who signs off (manager, IT/security, privacy/legal, HR if it affects employees).
Turn this list of tools into an approved-tools register table: tool, plan/account type, approved uses, allowed data classes, owner, next review date: [list]
Summarise this policy as 8 do's and don'ts that fit on one screen, for all staff: [paste policy]
Illustrative example
You paste
40-person accounting firm in Ontario; staff use ChatGPT and Copilot; client financial data is confidential.
A good answer looks like
5. Data rules: Public information may be used in any approved tool. Client financial data may only be used in tools marked 'Confidential' in the register, under a business plan where the vendor does not train on our inputs [CHECK WITH LEGAL]…
Illustrative example written for this guide, not output captured from a specific tool.
Safe practice
- Have a lawyer or privacy professional review the final policy; AI drafts are not legal advice (see Giving legal advice to a client).
Which tool to pick, and why
- Claude (incl. Claude Code): Team and Enterprise plans list admin features such as SSO, central billing and admin controls for connectors, and model training is 'None by default' on Team and Enterprise (claude.com/pricing, checked Oct 23, 2026). Pro, for drafting the policy, is US$20/month or US$17/month billed annually. Free plan can do this Visit official website ↗
- ChatGPT (incl. Work & Codex): OpenAI's pricing page describes Business and Enterprise plans for organisations (Business from 2 users) and lists encryption in transit and at rest (chatgpt.com/pricing, checked Oct 23, 2026). Free plan can do this Visit official website ↗
- Google Gemini: Google's plans page says Gemini Enterprise business data, including prompts and outputs, isn't used to train Google models (gemini.google/subscriptions, checked Oct 23, 2026). Free plan can do this Visit official website ↗
- Microsoft Copilot: Microsoft says Copilot is built on its existing security, privacy and compliance commitments, relevant if your company already runs Microsoft 365 (microsoft.com, checked Oct 23, 2026). Free plan can do this Visit official website ↗
Reasons come from the capabilities, free-plan details and prices we've checked on official pages (separate from any dated trial above). Sources: claude.com/pricing · chatgpt.com/pricing · gemini.google/subscriptions/ · www.microsoft.com/en-us/microsoft-copilot/for-individuals
Common mistakes
- Copying a template without adapting it to your laws, contracts and data.
- Banning all AI, which pushes staff to personal accounts you can't see.
- Approving a tool once and never re-checking when its terms change.
- No owner and no way to ask for a new tool.
Privacy tip
When drafting, describe your company in general terms; you don't need to paste client contracts or internal security details into a chat to get a useful first draft.
Claude (incl. Claude Code) setting · ChatGPT (incl. Work & Codex) setting · Google Gemini setting · Microsoft Copilot setting · Sources: privacy.claude.com/en/articles/12109829-how-do-i-change-my-model-improvement-privacy-settings · help.openai.com/en/articles/7730893-data-controls-in-chatgpt · support.google.com/gemini/answer/13594961 · support.microsoft.com/en-us/privacy/microsoft-copilot/privacy-controls
See also
- Compliance checklist: turn the policy into a compliance checklist
- Security policy drafting: broader security policy drafting
- Plan change management and communications: roll out the policy as a change
- Create an AI tool approval process: focused AI tool approval process and request form