AI Use Policy
A starting-point, organization-wide policy on acceptable AI use, data handling, human accountability and disclosure, to adapt with your counsel.
This is a template, not legal advice. It is a starting point written for organizations deploying AI teammates and other AI tools. Adapt it to your jurisdiction, industry and existing policies with your own legal counsel before adopting it. Regulatory references are for orientation only.
What this is for
Most employees are already using AI, approved or not. A clear policy tells everyone what is allowed, what is not, how to handle data, and who is accountable for AI-assisted work. It covers all AI tools, not just Clones, because the risks (data leakage, unchecked output, undisclosed use) are the same. The Clone Governance Charter sets rules for each Clone; this policy sets rules for every person.
When to use it
Adapt it in the Week 3 governance design session, alongside the Clone Governance Charter. Publish it before the first department pilot. Review it annually, and whenever a new AI tool category or regulation comes into scope.
Who's involved
- InfoSec lead: owns the policy.
- Legal counsel: reviews and adapts for jurisdiction and industry.
- HR lead: aligns with the code of conduct and disciplinary process.
- Executive sponsor: approves and announces it.
- GRC Specialist and GRC Clone: draft, track acknowledgments, keep the approved tools list current.
How to fill it in
- Replace every
[bracketed]item with your organization's details. - Compare each section with existing policies (acceptable use, data classification, records retention, code of conduct). Reference them rather than duplicating them.
- Build the approved tools list (Section 6) from the pre-read questionnaire's AI inventory.
- Agree the data rules table (Section 5) with InfoSec, using your own classification levels.
- Send to legal counsel for review. Record changes.
- Approve, publish, and require acknowledgment from every employee and contractor who uses AI tools.
The template
[Organization] AI Use Policy
Version: [1.0] · Effective date: [date] · Owner: [InfoSec lead] · Approved by: [names] · Next review: [date]
1. Purpose
This policy explains how people at [Organization] may use artificial intelligence tools, including AI teammates ("Clones"), AI assistants and AI features inside other software. Its goals are to help people use AI productively, to protect our data and our customers, and to keep people accountable for decisions.
2. Scope
This policy applies to all employees, contractors and temporary staff, and to any AI tool used for [Organization] work, on any device, whether provided by [Organization] or not.
3. Principles
- People stay accountable. You are responsible for any work you produce or approve with AI help, as if you had done it yourself.
- AI adds capacity. We use AI to give people time back for higher-value work, not to remove human judgment from decisions that need it.
- Approved tools only. Use AI tools on the approved list for company work.
- Protect data. Only put data into an AI tool if that tool is approved for that data class.
- Check before you rely. AI output can be wrong, incomplete or invented. Verify it in proportion to the stakes.
- Be open about AI involvement where it matters to the people receiving the work.
4. Acceptable and unacceptable use
| Generally acceptable | Requires approval | Not permitted |
|---|---|---|
| Drafting, summarizing and researching internal work | New AI tools or new connections to company systems | Using unapproved AI tools with Confidential or Restricted data |
| Using approved Clones for their documented workflows | Customer-facing AI use (chat, phone, SMS, email) | Using AI to make final decisions with legal or similarly significant effect on a person without human review |
| Meeting notes and action items, with participants informed | Using AI on Restricted data | Generating content that is deceptive, discriminatory, harassing or infringing |
| Code assistance in approved tools, with normal code review | Changing a Clone's autonomy level (per the Clone Governance Charter) | Impersonating a real person, or presenting AI as a human when asked |
| Building custom skills or automations that act in company systems | Bypassing logging, approval gates or security controls |
5. Data handling rules
| Data class | [Approved enterprise AI tools, incl. Clones] | [Consumer or unapproved AI tools] |
|---|---|---|
| Public | Allowed | Allowed |
| Internal | Allowed | Not allowed |
| Confidential | Allowed within approved Clones and connections | Not allowed |
| Restricted (e.g. [PHI, payment card data, government IDs, special-category personal data]) | Only where InfoSec has approved the specific use | Never |
Additional rules:
- Do not paste credentials, API keys or passwords into any AI tool.
- Personal data must be handled in line with [privacy policy] and applicable data protection law.
- AI outputs that become business records follow [records retention policy].
6. Approved tools
The current approved list is maintained by [InfoSec] at [location]. Each entry states the tool, approved data classes, approved uses and owner.
| Tool | Approved data classes | Approved uses | Owner |
|---|---|---|---|
| [Clone Team Studio] | [Up to Confidential; Restricted where approved] | [Documented Clone workflows] | [IT lead] |
To request a new tool, submit [form or ticket]. InfoSec aims to respond within [10] business days.
7. Human oversight
- Clones and other AI agents act only at the autonomy level approved in the Clone Governance Charter.
- Anyone approving an AI action must review it as carefully as if a colleague had prepared it.
- If an AI tool or Clone does something unexpected, stop using it for that task and report it (Section 10).
8. Disclosure of AI involvement
- Tell customers when they are interacting with an AI teammate, where [law, contract or our customer commitments] require it, and whenever they ask.
- Note AI assistance in [regulated documents, external publications, legal filings, audit evidence] as required by [policy or regulator].
- Internal work does not need a label unless the recipient would reasonably rely on it being human-authored.
9. Intellectual property and confidentiality
- Do not enter third-party confidential information into AI tools unless our agreement with that party allows it.
- Review AI-generated content for third-party material before external publication.
- Work created with AI tools for [Organization] belongs to [Organization] under [employment or contractor agreement].
10. Reporting and incidents
Report any suspected misuse, data exposure or harmful AI output to [security contact or channel] immediately. Reports made in good faith will not lead to disciplinary action against the reporter.
11. Training
Everyone who uses approved AI tools completes [AI fluency training] within [30] days of access. Champions and admins complete additional training.
12. Compliance
Breaches of this policy are handled under [disciplinary policy]. [Organization] may monitor AI tool use in line with [monitoring policy] and applicable law.
13. Regulatory references (orientation only, not legal advice)
[Counsel to confirm which apply: e.g. GDPR, the EU AI Act, HIPAA, sector rules, state AI and privacy laws.]
14. Acknowledgment
I have read and understood the [Organization] AI Use Policy and agree to follow it.
Name: __________ Signature: __________ Date: ______
Rollout checklist
- Legal review complete and changes recorded
- HR has aligned with the code of conduct
- Approved tools list published
- Policy announced by the executive sponsor
- Acknowledgment collected from all AI tool users (target: 100% before access)
- Policy linked from the Clone onboarding and training curriculum
Good looks like
- Short and readable. If the policy needs a lawyer to follow it, people won't.
- Covers all AI tools, including the ones already in use without approval. A realistic path to approval reduces shadow AI.
- Data rules map to your own classification levels, not new ones.
- Accountability is clear: the person who uses or approves the output owns it.
- Common mistake: banning AI outright. It drives use underground and removes visibility.
How Clone helps
The GRC Clone drafts the policy from your existing policies and the pre-read, highlights conflicts with current documents, keeps the approved tools list current, and tracks acknowledgments. The Enablement Clone answers employee questions about the policy and folds it into training. In the portal, this is the AI Use Policy deliverable under Security & GRC; legal review and approval are recorded against it, and the annual review date is tracked.
Sign-off
| Role | Name | Date | Decision |
|---|---|---|---|
| InfoSec lead | Approved | ||
| Legal counsel | Reviewed | ||
| HR lead | Aligned | ||
| Executive sponsor | Approved for publication |