Malaysian marketing executive using a laptop chatbot alone at her office desk after hours

Shadow AI: Your Staff Are Already Using AI Without Telling You

Short answer: Shadow AI is staff using AI tools the company has not approved, usually on personal accounts, to get work done faster. It is already common: 71% of UK employees in a 2025 Microsoft-commissioned survey said they use unapproved AI tools at work.

The answer is not a ban. It is visibility, approved tools and clear ground rules.

Picture an office like yours this week.

Someone pastes work material into a personal AI account.

Not out of carelessness. They are trying to finish their work.

What shadow AI is

Shadow AI is any AI tool used for work without the company's knowledge or approval.

Personal chatbot accounts, free summarisers, image generators, translation apps.

The pattern is not new. Staff once brought their own spreadsheets and cloud drives before IT caught up.

AI is the fastest-moving version of that story.

What the research says

  • 71% of UK employees said they used unapproved "shadow AI" tools at work, in a survey of more than 2,000 workers commissioned by Microsoft and reported by Euronews in October 2025. Only 30% of those users were concerned about company or customer privacy.
  • Half of the US workforce uses AI tools at work without knowing whether it is allowed, and 44% admit to using AI improperly, in KPMG's 2025 global study on trust in AI. 58% rely on AI output without properly evaluating it, 57% say that reliance has led to mistakes, and 53% avoid disclosing their AI use.
  • 75% of global knowledge workers already use AI, according to the 2024 Microsoft and LinkedIn Work Trend Index, with employees bringing their own tools rather than waiting for company plans.

Read together, the picture is consistent across countries: usage is far ahead of policy, and people keep quiet about it.

Why staff go around the rules

Shadow AI need not be rebellion. It can reflect staff trying to get work done.

  • The work is real. Deadlines, inboxes, reports. The tool helps tonight, not after a procurement cycle.
  • The approved option is missing or slow. In the UK survey, over 40% of shadow AI users said they simply used what they already knew from personal life.
  • Nobody has said what is allowed. Half the US workforce in the KPMG study did not know whether their AI use was permitted.
  • Silence feels safer than asking. When the policy is unclear, staff can assume asking draws attention, so they stay quiet.

Why a ban backfires

A ban without an alternative risks moving usage rather than stopping it.

The tool can move from the work laptop to the personal phone. The company can lose the little visibility it had.

There is a possible second cost.

The people most likely to comply with a ban are the careful ones. They would lose a real productivity edge if others carry on quietly.

The real risks, and they are real

  • Data leaves the building. Client names, pricing, personal data and draft contracts can end up in tools with unknown data handling. For Malaysian teams, that sits badly with PDPA obligations and client confidentiality.
  • Unverifiable output gets trusted. Shadow use can come without a checking habit, so confident wrong answers can flow straight into work.
  • No support when it breaks. If a free tool changes its terms, leaks data or disappears, there is no fallback and often no record of what went into it.
  • The company cannot learn. When usage is hidden, nobody knows what actually works, so the official policy gets written in the dark.
    Malaysian team agreeing three ground rules for AI tools at a whiteboard in a meeting room

What leaders can do instead

  • Ask before you ban. A short anonymous survey tells you more than a policy draft. You cannot manage what you have not measured.
  • Approve a shortlist. Pick one or two tools for work use, check their business terms and data settings rather than assuming a paid tier alone protects the data, and say clearly what they may be used for.
  • Write three ground rules, not thirty. What data never goes in, what output always gets checked, and who to ask when unsure.
  • Make asking safe. The goal is that the first question about a new tool comes to you, not to a search engine.
    Infographic showing hidden AI use on personal devices moving to approved tools, ground rules and safe asking
  • Keep the verification habit. Every AI-assisted output gets checked by a named person before it leaves the desk. The data side of this is covered in our guide to using AI at work safely.

A picture from an ordinary office (illustrative, not a real company)

Picture a marketing executive, call her Mei Ling, pasting a long client brief into a personal AI account on a Sunday night: "Summarise this into five key points for tomorrow's meeting."

The brief contained client pricing. The clean summary that comes back quietly drops the one condition the client stressed.

Nobody compares it against the original, because the tool and the habit are invisible to the company.

A practiced reviewer would check. In a good course, learners run exactly this comparison on their own summaries and see what the tool left out.

Nothing goes wrong this week.

But her company now has a habit it cannot see, and a habit it cannot see is a risk it cannot manage.

The fix is not catching Mei Ling. The fix is making the approved path faster than the shadow one.

The Malaysian angle

If Malaysian workplaces follow the pattern these surveys found elsewhere, adoption here is likely running ahead of policy too.

That kind of gap is where shadow AI can grow.

Two local realities raise the stakes.

Client trust is the currency of most Malaysian service businesses, and personal data handling carries PDPA obligations. A well-meaning paste into the wrong tool can put both at risk before anyone notices.

The leadership read: shadow AI is a signal before it is a security problem. When capable staff quietly route around official tools, they may be saying the official tools are behind.

Listen first, then set the rules.

Frequently asked questions

What counts as shadow AI?

Any AI tool used for work that the company has not approved: personal chatbot accounts, free online summarisers, AI features inside personal apps. If IT or management does not know it is being used for work, it is shadow AI.

Is shadow AI illegal in Malaysia?

Shadow AI is an approval and visibility issue, not a legal category. Whether a particular use is lawful depends on the data involved, the purpose and the rules that apply to your industry, including Malaysia's PDPA for personal data and any client confidentiality terms. For a specific situation, get qualified legal advice.

Should we just block AI sites on the office network?

Blocking without an alternative can push usage to personal phones and home laptops, where you have even less visibility. A shortlist of approved tools plus clear ground rules works better than a wall.

How do we find out how much shadow AI we already have?

Ask, anonymously. A five-question survey about which tools people use, and what for, opens the conversation instead of starting a crackdown.

The organisations that handle shadow AI well are rarely the ones with the strictest bans. They are the ones whose approved path is faster than the shadow one.

Michael Cheong runs hands-on AI training for Malaysian workplace teams, built around exactly these habits: approved tools, clear ground rules and a checking culture that sticks. Details are on the AI Trainer for Workplace Productivity page.

Similar Posts

Leave a Reply

Your email address will not be published. Required fields are marked *