AI policy template
for an acceptable-use workshop
Draft the policy. See what is still open.
Try the templateThe film is unavailable. You can read what it shows below.
What the film shows, in words
- Draft the policy. See what is still open.
- 3 allowed uses. 4 kinds of data banned.
- Review: anything shared outside. Disclosure: when material. Training before use.
- 2 decisions still open. Which tools are approved, and who approves new ones. Where staff report errors, leaks or harmful outputs.
- Allow applicant screening: 3 decisions open, and a watch point. Screening applicants is a high-risk use.
- A draft for the workshop. The decisions stay yours.
The demo: the worked example, one decision at a time
ExampleA draft for discussion, not legal advice.
Seven clauses written. Two holes.
Worked example: an NGO allowing drafting and translation, with strict data rules.
3 allowed uses, 4 data categories banned, 2 decisions still open: clause 4, Approved tools and clause 8, Incidents.
Example NGO: AI acceptable-use policy (draft for discussion)
1Purpose
This policy sets out how staff may use AI tools safely, lawfully and in line with our mission.
2Allowed uses
- Drafting and editing text
- Summarising documents
- Translation
3Data that must never be entered
- Personal data about clients or beneficiaries
- Health or other sensitive data
- Bank and payment details
- Passwords, keys and credentials
4Approved tools
[Decision open]
Decision 01Which tools are approved, and who approves new ones.
5Human review
A person reviews every AI output before it is shared outside the organisation.
6Disclosure
We disclose AI use when it materially shaped an output.
7Accountability and training
- The Director of Operations is accountable for this policy.
- Staff complete AI training before using approved tools.
8Incidents
[Decision open]
Decision 02Where staff report errors, leaks or harmful outputs.
9Review
This policy is reviewed every twelve months, or sooner when law or tools change.
Now your policy, written on the sheet.
Choose what your organisation allows, what data must never go into AI tools, who is accountable and how outputs are reviewed. The template drafts policy clauses and lists every decision still open, ready for a leadership workshop.
Make the choices; open decisions are listed, not guessed. Leave a clause undecided and the sheet keeps a hole there.
A draft for discussion, not legal advice. Have it reviewed against the laws that apply to you before adoption.
Read the draft as plain text
Example NGO: AI acceptable-use policy (draft for discussion)
1. Purpose
This policy sets out how staff may use AI tools safely, lawfully and in line with our mission.
2. Allowed uses
- Drafting and editing text
- Summarising documents
- Translation
3. Data that must never be entered
- Personal data about clients or beneficiaries
- Health or other sensitive data
- Bank and payment details
- Passwords, keys and credentials
4. Approved tools
[Decision open]
5. Human review
A person reviews every AI output before it is shared outside the organisation.
6. Disclosure
We disclose AI use when it materially shaped an output.
7. Accountability and training
- The Director of Operations is accountable for this policy.
- Staff complete AI training before using approved tools.
8. Incidents
[Decision open]
9. Review
This policy is reviewed every twelve months, or sooner when law or tools change.
Decisions for the workshop
- 01Which tools are approved, and who approves new ones.
- 02Where staff report errors, leaks or harmful outputs.
Watch points
- Strength: The core data protections are in place.
A draft for discussion, not legal advice. Have it reviewed against the laws that apply to you before adoption.
One page for each office, from the clauses you chose.
Every rule on a card is a line of a clause selected on the sheet, word for word, under its clause number. A choice still open is listed as open, never as a rule.
No office is named yet. This is the card as it would read today for one office. Name your offices in part B of the sheet and each gets its own.
- Draft stays on the card until you enter your own approval record and no choice is left open.
- No authority. A card is a summary to pin up. It gives no one a power to approve, sign or buy.
- Prints alone, on one A4 page. Nothing is sent anywhere.
Example NGO
Specimen · Draft
AI safe-use cardYour office or unit
Draft. No approval is recorded.
You may use AI forclause 2
- Drafting and editing text
- Summarising documents
- Translation
Never put into an AI toolclause 3
- Personal data about clients or beneficiaries
- Health or other sensitive data
- Bank and payment details
- Passwords, keys and credentials
- 5Have a person checkclause 5
- A person reviews every AI output before it is shared outside the organisation.
- 6Say when AI helpedclause 6
- We disclose AI use when it materially shaped an output.
- 7Before you startclause 7
- The Director of Operations is accountable for this policy.
- Staff complete AI training before using approved tools.
Still open: not rules yet
- Clause 4, Approved tools
- Clause 8, Incidents
- 1Purposefixed text
- 2Allowed useswritten
- 3Data that must never be enteredwritten
- 4Approved toolsopen
- 5Human reviewwritten
- 6Disclosurewritten
- 7Accountability and trainingwritten
- 8Incidentsopen
- 9Reviewfixed text
- Allowed uses
- 3
- Banned data
- 4
- Open decisions
- 2
Your draft
Draft policy for a workshop, by email once delivery is connected.
- The nine clauses, as you left them
- Every decision still open, in the tool’s order
- The watch points, for a qualified reviewer
Tool: AI policy template Organisation: Example NGO Result: 3 uses · 2 open decisions Status: Draft. No approval is recorded. Agreed rules: 11 lines. Open choices: 2. Authority: This document confers no authority by itself. It gives no person or office a power to approve, sign, buy or decide. Authority comes only from your organisation’s own approving body and its records. Worked out for: Context not set · method 1.1 Request: please facilitate a policy workshop (SVC-028 / SVC-030).Send as a brief
How the draft is made, and what it is not.
The rules are published so you can check them. A draft for discussion, not legal advice. Have it reviewed against the laws that apply to you before adoption.
Method
- 01
Each choice maps to a standard clause; anything you leave undecided is marked “[Decision open]” and listed for the workshop.
- 02
High-risk uses (hiring, public-facing chat) add their own open questions.
- 03
Missing core data bans are flagged.
- 04
Every line is marked as an agreed rule (a choice made on the sheet), an open choice (left for the workshop) or template wording. Further clauses are proposed practice and become rules only when you select them.
- 05
A safe-use card for an office repeats, word for word, the lines of the clauses you selected, each under its clause number. An open choice is listed on the card as open, never as a rule.
- 06
The document stays marked “Draft” until you enter your own approval record and no choice is left open. The tool does not check that record.
What it cannot tell you
- A starting draft, not legal advice or assurance.
- Sector rules (health, finance, education) may add requirements.
- A policy only works with training, monitoring and an owner who enforces it.
- The document and its cards confer no authority by themselves. A role you name is your own words: no power to approve, sign or buy follows from it.
- No national law or guidance module has been reviewed for this tool. Name a country and the result says which coverage is missing.
Quality ruleMark unresolved policy choices; require principal review.
When the open decisions need a workshop.
The tool lists what leadership still has to decide. Deciding it is work for people in a room.
- Service · 03 Training & capacity buildingExecutive AI literacy & board AI-governance day↗
- Service · 03 Training & capacity buildingEU AI Act Article 4 AI-literacy programme↗
-
Michael GagakumaTool owner · consultations from USD 90Book Michael -
Kenneth GagakumaTool owner · consultations from USD 90Book Kenneth
