Guide · Malaysia & Singapore
Shadow AI: How to Find It and Stop It
Your staff are already using AI tools you never approved. This guide shows how to find them in 30 days without a witch-hunt, and how to move people onto an approved tool, using our free shadow AI discovery procedure (AGK-PRC-07 v1.0).
Short answer
What is shadow AI?
Shadow AI is any AI tool staff use for work without the organisation's approval: a personal ChatGPT or Claude account, a browser extension, or an AI feature switched on inside a SaaS app. The risk is that company or personal data goes into a tool nobody has checked, and nobody can show what was shared.
Why it matters
What are the risks of shadow AI?
The tool is rarely the problem. The problem is that the company cannot see it, so it cannot set rules for it, check its terms, or answer for what went into it.
- 01
Company data leaves your control
Staff paste customer lists, contracts, pricing or source code into a tool nobody has checked. Personal and free accounts sit outside your contract, your retention settings and your single sign-on.
- 02
No record of what was shared
With no company account there are no audit logs. When a client, an auditor or a regulator asks what went into the tool, nobody can show it.
- 03
A personal data breach you cannot assess
If personal data went in, it may be a personal data breach. Malaysia's PDPA, as amended in 2024, sets a 72-hour notification clock, and you cannot assess a breach in a tool you did not know about.
Same tool, two answers
Is ChatGPT shadow AI?
It depends on the account, not the tool. ChatGPT, Claude, Copilot or Gemini through a company account on the approved AI tool list is governed AI. The same tool on a personal account, used for work, is shadow AI.
| Company account on the approved list | Personal account used for work | |
|---|---|---|
| Is it shadow AI? | Company account on the approved listNo. It is governed AI. | Personal account used for workYes, even if the same tool is approved. |
| Sign-in | Company account on the approved listCompany identity and single sign-on | Personal account used for workPersonal email address |
| Contract and data use | Company account on the approved listCompany agreement, configured to the baseline | Personal account used for workConsumer terms the company never accepted |
| Audit logs | Company account on the approved listAvailable to the administrator | Personal account used for workNone the company can see |
| Leaver | Company account on the approved listAccess removed with the company account | Personal account used for workThe history stays with the person |
| What data may go in | Company account on the approved listUp to the limits of the data × AI tool matrix | Personal account used for workPublic information only (AGK-POL-02-C02) |
Rule: AGK-POL-02 v1.1, control C02. See the data × AI tool matrix for what may go into an approved tool.
Discovery · AGK-PRC-07 v1.0
How to find shadow AI
Use as many sources as you have. Two or three together give a reasonable picture; no single source finds everything. A company without network tools can still cover the first four.
| Source | What it finds | Works for a small company? |
|---|---|---|
| Staff survey | What it findsWhat people use, for what and with what data, including personal devices and accounts no log can see | Works for a small company?Yes |
| Expense and corporate card data | What it findsAI subscriptions bought by individuals or teams (search vendor names, "AI", "GPT", "assistant") | Works for a small company?Yes |
| OAuth grants in your productivity suite | What it findsThird-party AI apps given access to mail, files or calendars, and the scopes granted | Works for a small company?Yes |
| AI features inside existing SaaS | What it findsAI switched on by default or by admins in CRM, HR, finance, helpdesk, design and meeting tools | Works for a small company?Yes |
| Browser extension inventory | What it findsAI extensions and sidebar assistants on managed browsers | Works for a small company?If devices are managed |
| DNS or proxy logs | What it findsVisits to known AI domains, by device or user, over the last 90 days | Works for a small company?If available |
| SaaS discovery, CASB or secure web gateway | What it findsSanctioned and unsanctioned AI web apps, users, volumes and uploads | Works for a small company?If available |
Run a 30-day no-blame amnesty
Logs only show company devices. People will tell you about personal accounts and phones only if they will not be punished for it, so announce the amnesty with the survey: anything declared in 30 days is moved to an approved tool, not disciplined.
Record it in the AI system register
Every tool you find goes into the AI system register, with an owner and the data it touches. The register is how you prove later what AI the company runs.
The staff survey, in ten questions
Anonymous by default and under five minutes. It asks which AI tools people used for work in the last three months, personal or company account, what for, how often, what kind of information went in, any AI browser extensions or meeting note-takers, any AI connected to work email or files, AI features they noticed in existing software, what they would like AI help with, and what worries them. The full wording is in the procedure.
Stop it
How to stop shadow AI
Remove the reason for it, in this order. Blocking first, before staff have an alternative, moves the problem somewhere you cannot see it.
Step 1
Give staff an approved tool
Most shadow AI is staff trying to work faster. Approve at least one AI tool through company accounts and single sign-on, so the safe route is also the easy one. The survey tells you which tasks it has to cover.
Compare ChatGPT plansStep 2
Publish the acceptable use policy
Say which tools are approved, what data may go into which tool, and that no non-Public company information goes into consumer AI, including personal accounts on approved tools. Every user signs before access.
AI acceptable use policy templateStep 3
Block personal accounts on company devices
Once the approved tool is live, block personal AI accounts and unapproved AI services on company devices and networks. Blocking before an alternative exists only pushes use onto personal phones.
Shadow AI discovery procedure
Checklist
A 30-day shadow AI sprint
One month from announcement to a closed amnesty. The procedure carries the owners and the RACI; this is the order of work.
Days 1–7
Announce and ask
- Name an owner for discovery (usually the AI Governance Lead) and get a sponsor's sign-off.
- Announce a 30-day no-blame amnesty for any AI tool staff declare themselves.
- Send the anonymous staff survey with the amnesty message; keep it under 5 minutes.
- Ask Finance for expense and corporate card lines that mention AI vendors.
Days 8–14
Look at the systems
- Pull OAuth grants for AI apps from your productivity suite.
- List AI features switched on inside the SaaS you already use.
- Where you have them, check DNS, proxy or SaaS discovery logs and the browser extension inventory.
- Chase survey replies; the survey is the only source that sees personal devices.
Days 15–21
Classify and record
- Classify every finding against the AI data handling standard and the tool tiers.
- Record every tool in the AI system register, with an owner and the data it touches.
- Decide the approved tool for the tasks staff named most often.
- Log anything where personal data went into an unapproved tool as an incident and assess it.
Days 22–30
Switch over and close the door
- Move staff from personal accounts to the approved tool, with a short how-to.
- Publish the acceptable use policy and collect signed acknowledgements.
- Block personal AI accounts on company devices once the approved tool is live.
- Close the amnesty and schedule a light discovery every quarter.
Free template · AGK-PRC-07 v1.0
The shadow AI discovery procedure template
A Word procedure you adapt: triggers, discovery sources, how to classify findings, the amnesty, the move from consumer to approved tools, workflow, controls, RACI and records. Highlighted [values] are the ones you set.
AI Discovery & Shadow AI Procedure
- Purpose
- Scope
- Triggers
- Discovery sources
- Classifying findings
- Amnesty
- Migration path from consumer (T0) to approved tools
- Workflow
- Controls
- RACI
- Records
- Related documents
- Framework alignment
Free AI Governance Lite pack
The documents an organisation starting out needs to govern AI tools in its first 90 days. Word and Excel, ready to edit.
- Lite Quick Start and Control Checklist AGK-001 · Word
- AI Governance Framework and Document Architecture AGK-000 · Word
- AI Acceptable Use Policy AGK-POL-02 · Word
- AI Data Handling Standard AGK-STD-A02 · Word
- AI Discovery and Shadow AI Procedure AGK-PRC-07 · Word
- AI Adoption Roadmap and Use-Case Prioritisation Guide AGK-GDL-C02 · Word
- AI System Register AGK-TPL-01 · ExcelPreview AI System Register
- AI Use Case Intake Form AGK-TPL-03 · Word
- AI Incident, Nonconformity and CAPA Log AGK-TPL-06 · ExcelPreview AI Incident, Nonconformity and CAPA Log
- AI Acceptable Use Acknowledgement AGK-TPL-07 · Word
- Pengakuan Penggunaan AI (Bahasa Melayu) AGK-TPL-07-BM · Word
- AI Use-Case Prioritisation and Benefits Tracker AGK-TPL-12 · ExcelPreview AI Use-Case Prioritisation and Benefits Tracker
Template release v1.0 · ZIP · 990 KB
Download AI Discovery and Shadow AI Procedure and the Lite pack
Templates for governance purposes, not legal advice.
AI Governance Hub
AI Governance Hub: the AI risk management platform for your risk team
Templates go stale. The Hub keeps every AI system registered, risk-assessed, owned and evidenced, and answers staff questions from your own policy. Launching soon: join the waitlist and we will contact you before it opens.
| Feature | Free | Member |
|---|---|---|
| Templates | 12 Lite documents | All 48 documents, with version history |
| Pre-read and tier check | ||
| 90-day planner with owners and evidence | ||
| Policy assistant: a pack tailored to your company | ||
| AI system and risk register, with a ready AI risk library | ||
| Control owner register with quarterly attestation and evidence | ||
| Connector, agent and AI vendor registries | ||
| Incidents, exceptions and review calendar | ||
| KRI dashboard and one-click board report | ||
| MCP connection: staff ask Claude, ChatGPT or Copilot and get answers from your own policy | ||
| Regulation and vendor-setting update alerts |
Member features are planned for launch and may change.
Enterprise · deploy in your environment · coming later
Run the same AI risk platform in your own cloud account (AWS, Azure or Google Cloud, with your own AI model account), or fully offline on your own servers with a private AI model. For organisations that cannot use shared cloud software. Pricing on request.
Want it set up with you? We run a governance setup sprint: owners, policy, register and an approved AI tool. Book an online meeting
RM 2,400
per organisation per year
Shadow AI FAQ
What is shadow AI?
Shadow AI is any AI tool staff use for work without the organisation's approval: a personal ChatGPT or Claude account, a browser extension, or an AI feature switched on inside a SaaS app. The risk is that company or personal data goes into a tool nobody has checked, and nobody can show what was shared.
Is ChatGPT shadow AI?
It depends on the account. ChatGPT through a company account on the approved AI tool list is governed AI. The same tool on a personal account, used for work, is shadow AI. Under the kit's acceptable use policy, no non-Public company information may go into consumer AI, including personal accounts on approved tools.
What are the risks of shadow AI?
Company data leaves your control, there is no record of what was shared, and personal data may have gone into a tool you never assessed. In Malaysia, the PDPA as amended in 2024 sets a 72-hour personal data breach notification clock, which is hard to meet for a tool you did not know about.
How do you find shadow AI?
Combine sources, because no single one finds everything: an anonymous staff survey, expense and card data, OAuth grants in your productivity suite, AI features inside existing SaaS, and network or browser logs where you have them. Pair it with a 30-day no-blame amnesty and record every tool in the AI system register.
How do you stop shadow AI?
Remove the reason for it. Give staff an approved AI tool through company accounts, publish an acceptable use policy that says what data may go where, then block personal AI accounts on company devices. Repeat a light discovery every quarter, because vendors keep switching on new AI features.
Should we ban AI tools to stop shadow AI?
A ban on its own usually moves AI use onto personal phones, where you see even less. The procedure in the kit approves a tool first, moves staff onto it, and only then blocks personal accounts on company devices. Staff who have a sanctioned route have little reason to work around it.
What is the difference between shadow AI and shadow IT?
Shadow AI is the AI part of shadow IT: technology used for work without approval. The difference is what flows into it. Staff paste documents, customer data and code straight into AI prompts, and AI features appear inside software you already approved, so discovery has to look inside approved apps too.
Part of the AI governance guide for Malaysia and Singapore. See also the AI acceptable use policy template and the template library.
