Approved tools
Names the AI tools staff may use, the accounts to use them with, and the route for requesting a new one. Anything not on the list needs a request first.
AI use case intakeThe staff-facing rules for AI: what the policy covers, how it differs from the organization’s AI policy, and how to write one people can follow.
An AI acceptable use policy is the document that tells employees and contractors which AI tools they may use, for which tasks, with which data, and what they must check before relying on the output.
It is the staff-facing layer of AI governance, shorter and more specific than the organization’s overall AI policy.
Names the AI tools staff may use, the accounts to use them with, and the route for requesting a new one. Anything not on the list needs a request first.
AI use case intakeStates which classes of data may go into which tools. Typical never-enter items are customer personal data, credentials and secrets, unreleased financials and source code, unless a tool is approved for them.
Requires a person to check AI output before relying on it, and sets a higher bar, human review and a record, for output that informs decisions about people.
Human oversightLists uses that are off limits whatever the tool, starting with the law. EU AI Act Article 5 bans practices such as social scoring and inferring emotions in the workplace.
Article 5| AI acceptable use policy | AI policy | |
|---|---|---|
| Audience | Every employee and contractor | Leadership, risk, legal, engineering and procurement |
| Scope | Day-to-day use of AI tools | How AI is built, bought, assessed, overseen and retired |
| Reference | NIST AI 600‑1, actions GV‑1.4‑002 and GV‑3.2‑003 | ISO/IEC 42001 clause 5.2; NIST AI RMF GOVERN 1.2 |
| Form | A few pages of plain rules with examples | Principles, roles, risk tolerance and required processes |
| Enforced through | Training, access controls and monitoring | Assessments, approvals, audit and management review |
Write rules for the AI tools people already use, not a hypothetical list. The inventory and shadow AI findings tell you which those are.
Reuse your existing data classification, so staff apply labels they already know.
Publish the approved list and a quick way to ask for an addition. A policy with no route to yes produces workarounds.
Article 4 asks for measures that fit staff knowledge and context of use. It does not require guaranteeing any specific level of literacy.
NIST action GV‑6.1‑010 suggests updating acceptable use policies to address proprietary and open-source generative AI technologies and data.
Straight answers, kept current. Last updated .
Ask us anything →No law requires a document with that title. The EU AI Act does require providers and deployers to take measures to support staff AI literacy (Article 4), and requires deployers of high-risk systems to use them in line with the instructions for use (Article 26(1)). A written policy is the usual way to show both. ISO/IEC 42001 requires an AI policy in clause 5.2.
It depends on the tool’s terms and your data classification, so the policy should answer by data class. Common entries on the never list are personal data about customers and employees, passwords, keys and other secrets, unreleased financial results, source code, and anything covered by a confidentiality agreement.
Pair every rule with a control or a record. Training and signed acknowledgment show that staff know the rules. Single sign-on limits access to approved tools. Monitoring for unapproved AI use shows where the policy is being bypassed, and a fast request route removes the reason to bypass it. Review the findings on a set schedule.
Primary sources for the figures and frameworks cited on this page. Last verified .
See how Cranium AI discovers, observes, governs, secures and proves every AI model, agent and vendor in your enterprise.
Cookie Policy · Privacy Policy
Global Privacy Control detected. Opt-out request honored: marketing tags stay off in this browser.