Do managers know their staff use unapproved AI tools?

Usually, yes. In a Cybernews survey of over 1,000 US employees, 59% admitted using unapproved AI tools at work — and among those, 57% said their direct manager actively supports the practice while another 16% said their manager simply doesn't care. Among executives and senior managers in the same survey, 93% said they use unapproved AI tools themselves.

What kind of company data do employees put into public AI tools?

PagerDuty's June 2026 survey found 88% of office professionals had shared work-related information with public tools like ChatGPT, Claude or Gemini: 43% shared emails and correspondence, 40% shared meeting notes or summaries, 34% entered customer data, and 31% put in financial information or confidential company documents.

how many employees secretly use ai tools at work

Q: How many employees secretly use AI tools at work?

About 31% of generative AI users say they use it at work without their employer knowing, according to Deloitte's September 2026 survey of 25,000 UK workers. A narrower PagerDuty survey of 1,250 office professionals at large companies found 66% had used an AI tool at work they believed was against company policy. Both figures are self-reported, so they are floors rather than ceilings.

Q: Is shadow AI actually causing damage, or is it a vendor talking point?

IBM's 2026 Cost of a Data Breach report, based on 602 breached organizations, found shadow AI involved in 43% of breaches, up from 20% the year before. Those breaches averaged $5.39 million against a global average of $4.99 million. Over the same period the share of organizations requiring approval before deploying AI fell from 45% to 38%.

Roughly one in three. 31% of generative AI users say they use it at work without their employer knowing, per Deloitte’s survey of 25,000 UK workers published September 16, 2026. Widen “secretly” to “used a tool at work you believed was against policy” and it’s 66%. Both are self-reported, so read them as floors.

The gap between those two numbers is the whole story.

so is it 31% or 66%?

They measure different things, and neither one is wrong.

Deloitte’s 31% is the clean version of the question: of people who use generative AI for work, how many are doing it where the employer can’t see? That survey is the biggest of its kind in the UK — 25,000 workers aged 18 to 70, fielded by Ipsos between May 7 and June 10, 2026. In the same data, 63% of UK working adults say they knowingly use generative AI for work and 24% use it every day. So the hidden third sits on top of a very large base.

PagerDuty’s 66% is the guilty-conscience version: have you ever used an AI tool at work when you thought it wasn’t allowed? That survey, run by Wakefield Research and published June 11, 2026, polled 1,250 office professionals in non-IT roles, all at companies with over $500 million in revenue, across the US, UK, Australia and Japan. Bigger companies, more rules, more rule-breaking.

For the denominator, Gallup’s Q2 2026 workplace survey of 22,573 employed US adults found 52% use AI in their job at least occasionally, 30% weekly or more, and 15% daily. Half the workforce is in play. A third of them are off the books.

do their managers actually know?

This is where the “rogue employee” framing falls apart.

A Cybernews survey of more than 1,000 US employees, fielded in August 2025, found 59% using unapproved AI tools. Among those people: 57% said their direct manager supports it, and 16% said their manager doesn’t care. That’s roughly three in four operating with their boss’s blessing or indifference.

And the bosses are worse. 93% of the executives and senior managers in that survey admitted using unauthorized AI tools themselves. PagerDuty found 81% of office workers believe leadership operates under a different set of rules, which turns out to be an accurate read of the situation rather than cynicism.

So “secret” rarely means secret from your manager. It means secret from whoever would have to write it down.

what are people putting into these tools?

PagerDuty asked, and the answer is: everything. 88% of office professionals have shared work-related information with public tools like ChatGPT, Claude or Gemini. The breakdown — 43% shared emails and correspondence, 40% meeting notes or summaries, 34% customer data, 31% financial information or confidential company documents.

Nearly a third of office workers have pasted confidential documents into a chat window. That number comes from people willing to admit it on a survey.

why do employees hide it?

Fear, mostly, and it’s specific. Deloitte found 64% of weekly AI users worry their manager will decide the AI can do their job. Nearly a quarter see a stigma attached to using it at work at all. Around half of the people using AI at work have had no formal training from their employer, and 65% report a lack of convincing leadership on how it should be used.

Then they pay for it themselves. 17% of generative AI users buy at least one tool out of their own pocket for work — Deloitte puts the UK total at £958 million a year. People are expensing nothing, hiding the receipt, and doing the job faster.

The incentive to confess is zero. 77% of PagerDuty’s respondents say their company’s AI restrictions are limiting their professional growth, and 75% would be likely to move to an employer offering better AI skills development. Punishing disclosure while competitors advertise AI training is a strategy with one outcome. It’s the same mechanism that makes AI adoption mandates backfire, running in reverse.

is shadow AI actually dangerous, or is that a vendor pitch?

Both, but the damage is measurable. IBM’s 2026 Cost of a Data Breach report, published July 29 and built on 602 breached organizations across 16 countries, found shadow AI involved in 43% of breaches — up from 20% the year before. Those breaches averaged $5.39 million, against a global average of $4.99 million.

The part nobody wants to put on a slide: governance went backwards while that was happening. Organizations with strict approval processes before deploying AI fell from 45% to 38%. Breached organizations with an AI policy in place: 32%, down from 37%. Only 40% applied access controls to their AI models and data at all.

the real split

Everyone frames this as companies versus rule-breakers. The numbers say otherwise. Employees use AI, their managers nod along, the executives use more of it than anyone, and 68% of breached companies had no AI governance policy to break in the first place.

The divide isn’t between companies with shadow AI and companies without it. It’s between the ones who have written something down and the ones who haven’t looked yet.