AI Services Companies

Shadow AI Usage Statistics 2026

Published November 9, 2026

Our analysts compared 18 published surveys and security telemetry reports released between November 2023 and September 2026, including KPMG's survey of 1,019 US respondents, PagerDuty's Shadow AI Survey of 1,250 office professionals, Netskope's Cloud and Threat Report 2026, and IBM's 2025 Cost of a Data Breach Report, to show how many employees use unapproved AI tools at work as of Q4 2026, what they share with those tools, and what it costs when it goes wrong.

Key Findings

  • 44% of US workers knowingly use AI at work in ways their employer has not authorized, and 46% have uploaded sensitive company information or intellectual property to public AI platforms (KPMG, April 2025).2
  • 66% of office professionals at companies with $500 million or more in revenue have used AI tools they believed were not permitted (PagerDuty, June 2026), and 45% of workers at enterprises with 1,000 or more employees used unsanctioned AI in the past 30 days (WalkMe, April 2026).3,5
  • The share of workplace generative AI users on personal AI accounts fell from 78% to 47% in the year to October 2025, while use of company-managed accounts rose from 25% to 62% (Netskope).9
  • 20% of the breached organizations IBM studied had a breach linked to shadow AI, which added as much as $670,000 to the average breach cost; breaches involving shadow AI averaged $4.63 million against a $4.44 million global average.20,21
  • IT and telecoms workers report the highest shadow AI use at 73%, followed by finance at 59%, while healthcare workers report 41% and education is lowest at 38% (TrustedTech, September 2026).15

Shadow AI is the use of AI tools at work that the employer's IT or security team has not approved, often a free or personal ChatGPT, Gemini, or Copilot account opened in a browser. Employees turn to these tools because they save time, and many of them first learned the tools at home. The risk for the employer is that emails, customer records, source code, and financial documents end up in services the company cannot see, control, or delete from. Shadow AI figures appear in a great deal of security and workplace coverage, yet the surveys behind them ask very different questions. We wanted one place where the main numbers appear side by side with their samples and dates.

Our analysts took employee survey results from KPMG and the University of Melbourne, PagerDuty and Wakefield Research, BlackFog, WalkMe, UpGuard, Deloitte UK, Software AG, Salesforce, Microsoft and LinkedIn, TrustedTech, Wolters Kluwer, and ManageEngine. We added security telemetry from Netskope and Cyberhaven, which measure what employees do on company networks, and breach-cost data from IBM's 2025 Cost of a Data Breach Report. Cisco and Gartner supplied figures on bans and on what security leaders see. US results come first wherever a study reports them separately, and all money is in US dollars.

In the following sections, we compare headline shadow AI rates across surveys, how the picture changed from 2023 to 2026, which AI tools employees use and which ones companies block, what data employees share, how use differs by industry and role, and what policies, bans, and breaches show about the cost of shadow AI.

Between 31% and 88% of Workers Use Unapproved AI, Depending on the Survey

Each survey words its shadow AI question a little differently. Some ask about tools a worker believed were banned, some about any tool IT never approved, and some only about the past 30 days, so we lined up the main 2025 and 2026 results with the sample behind each one.

Shadow AI Use Rates in Recent Surveys
SurveyFieldwork or releaseSampleMeasureShare
KPMG and University of MelbourneNov 2024 to Jan 20251,019 US respondentsUS workers using AI at work without knowing whether it is allowed50%
KPMG and University of MelbourneNov 2024 to Jan 20251,019 US respondentsUS workers knowingly using AI without authorization or improperly44%
PagerDuty and Wakefield ResearchReleased Jun 20261,250 office professionals at $500M+ revenue firms (500 in the US)Used AI tools they believed were not permitted66%
BlackFogReleased Jan 20262,000 workers at companies with 500+ employeesAdopted AI tools without employer approval49%
WalkMeReleased Apr 20262,050 workers at enterprises with 1,000+ employees, 14 countriesUsed unsanctioned AI tools in the past 30 days45%
UpGuardJul to Aug 20251,020 US and UK employeesUse unapproved AI tools81%
UpGuardAug 2025542 security leadersUse unapproved AI tools88%
Deloitte UKMay to Jun 2026UK workers who use GenAI for workUse GenAI at work without their employer knowing31%

Source: AI Services Companies analysis of KPMG, PagerDuty, BlackFog via CIO, WalkMe, UpGuard, and Deloitte UK survey results, conducted November 2026. Data: shadow-ai-usage-rates-by-survey.csv

Dot chart of shadow AI use rates by survey: KPMG US workers 50% unsure if allowed and 44% knowingly unauthorized, UpGuard employees 81% and security leaders 88%, BlackFog 49%, WalkMe 45% in the past 30 days, PagerDuty 66%, and Deloitte UK 31%

Source: AI Services Companies analysis of KPMG, UpGuard, BlackFog, WalkMe, PagerDuty, and Deloitte UK survey results, conducted November 2026.

Key Findings

  • Senior people break the rules too. 69% of presidents and C-suite members and 66% of directors and senior vice presidents in BlackFog's sample are comfortable with unapproved AI use when it speeds up work, and UpGuard found 88% of security leaders using unapproved tools.4,6
  • Older studies show the same pattern. Microsoft and LinkedIn found in 2024 that 78% of AI users brought their own AI tools to work, and Software AG found in late 2024 that about half of knowledge workers used shadow AI.13,14
  • Many workers do not know where the line is. Half of US workers in KPMG's study use AI without knowing whether it is allowed, and 34% of WalkMe's workers do not know which AI tools their employer approves, even though 86% of PagerDuty's respondents believe their company has an AI policy. A written policy that staff cannot interpret does little to bring these rates down.2,5,3

Personal AI Accounts Declined from 2024 to 2025 While AI Use Kept Growing

Survey questions changed from year to year, so we paired the annual survey readings with two telemetry series that measure the same thing each time: Netskope's count of which accounts employees use to reach AI apps, and Cyberhaven's measure of how much of the data sent to AI tools is sensitive.

Shadow AI Measures by Year, 2023 to 2026
YearMeasureValueSource
2023Generative AI users at work using tools without formal employer approvalMore than halfSalesforce (14,000+ workers, 14 countries)
2023Share of corporate data put into AI tools that is sensitive (March)10.7%Cyberhaven
2024AI users bringing their own AI tools to work78%Microsoft and LinkedIn (31,000 people, 31 countries)
2024ChatGPT use at work through non-corporate accounts (March)73.8%Cyberhaven
2024Share of corporate data put into AI tools that is sensitive (March)27.4%Cyberhaven
2024Knowledge workers using shadow AI (US, UK, and Germany)About halfSoftware AG (6,000 workers)
2024Generative AI users on personal AI accounts (year-earlier baseline)78%Netskope
2025Generative AI users on personal AI accounts (October)47%Netskope
2025Generative AI users on company-managed AI accounts (October)62%Netskope
2025Share of corporate data put into AI tools that is sensitive34.8%Cyberhaven
2025US workers knowingly using AI without authorization44%KPMG
2026Office professionals who used AI they believed was not permitted66%PagerDuty
2026Workers using unsanctioned AI in the past 30 days45%WalkMe

Source: AI Services Companies analysis of Salesforce, Cyberhaven 2024, Cyberhaven 2025, Microsoft and LinkedIn, Software AG via eWeek, Netskope, KPMG, PagerDuty, and WalkMe data, conducted November 2026. Data: shadow-ai-trend-2023-2026.csv

Line chart: generative AI users on personal accounts fell from 78% to 47% and company-managed account use rose from 25% to 62% from 2024 to 2025 (Netskope), while the sensitive share of data sent to AI tools rose from 10.7% in 2023 to 27.4% in 2024 and 34.8% in 2025 (Cyberhaven)

Source: Netskope Threat Labs, Cloud and Threat Report: 2026; Cyberhaven 2024 and 2025 AI adoption reports.

Key Findings

  • Netskope found that the number of people using SaaS generative AI apps tripled in the average organization over the year, and prompts rose sixfold from 3,000 to 18,000 a month. The share of users switching between personal and company accounts grew from 4% to 9%.9
  • Because users tripled while the personal-account share fell by about two-fifths, the number of people on personal AI accounts in the average organization still grew, by an estimated 1.8 times. Cyberhaven also measured AI usage frequency at work rising 4.6 times in one year and 61 times in two years.9,11
  • Company-managed accounts now cover most AI users, which gives security teams more visibility than they had in 2024. The same data points to more total shadow AI activity through 2025, since overall AI use grew faster than the personal-account share fell.

ChatGPT and Gemini Lead Workplace AI Use, and DeepSeek Is Among the Most Blocked

Employers and reporters both ask which AI apps employees are reaching for, so we compared the apps organizations see on their networks, the ones they block outright, and how much use of the biggest apps ran through personal accounts.

Generative AI Apps in Use and Blocked at Organizations
AppOrganizations where it is in use (2025)Organizations actively blocking it (2025)Work use through non-corporate accounts (March 2024)
ChatGPT77%n/a73.8%
Google Gemini69%n/a94.4%
Microsoft 365 Copilot52%n/an/a
Perplexity35%n/an/a
Grok28%n/an/a
ZeroGPTn/a45%n/a
DeepSeekn/a43%n/a

Source: AI Services Companies analysis of Netskope Cloud and Threat Report: 2026 and Cyberhaven, May 2024, conducted November 2026. Data: genai-apps-used-and-blocked.csv

Horizontal bar chart of generative AI apps: in use at 77% of organizations for ChatGPT, 69% Google Gemini, 52% Microsoft 365 Copilot, 35% Perplexity, and 28% Grok; actively blocked by 45% for ZeroGPT and 43% for DeepSeek

Source: Netskope Threat Labs, Cloud and Threat Report: 2026.

Key Findings

  • Google Gemini rose from 46% to 69% of organizations over the year and Perplexity from 23% to 35%, while ChatGPT averaged 77%. Grok, once one of the most frequently blocked apps, began gaining ground in April 2025 and reached 28%.9
  • 90% of organizations actively block at least one generative AI app, and the average organization blocks 10. Netskope tracked more than 1,600 generative AI apps, up from 317 a year earlier, yet the average organization uses 8, up from 6.9
  • Free versions dominate shadow use: 58% of BlackFog respondents who use non-approved tools rely on free versions, and 34% of workers at companies that allow AI still choose the free version of an approved tool. Paid enterprise plans typically let the company turn off model training on its data, while virtually all free tools use what they ingest to train their models.4

Most Office Staff Have Put Work Information into Public AI Tools

The main harm from shadow AI comes from what employees paste into it, so we gathered the survey results on which kinds of company information workers say they have shared, along with what security tools catch on company networks.

Company Information Employees Report Sharing with AI Tools
Information sharedShare of respondentsSurvey
Any work-related information, with public AI tools88%PagerDuty, 2026
Emails and other correspondence43%PagerDuty, 2026
Meeting notes or summaries40%PagerDuty, 2026
Customer data or information34%PagerDuty, 2026
Financial information or confidential documents and strategies31%PagerDuty, 2026
Enterprise research or datasets, in unsanctioned tools33%BlackFog, 2026
Employee data such as salary or performance tracking27%BlackFog, 2026
Company financial information23%BlackFog, 2026
Sensitive company information or intellectual property, to public AI platforms (US workers)46%KPMG, 2025
Used unsanctioned AI with confidential data in the past 30 days36%WalkMe, 2026
Privacy and security professionals who have entered non-public company information into GenAI tools48%Cisco, 2024
Privacy and security professionals who have entered employee information into GenAI tools45%Cisco, 2024

Source: AI Services Companies analysis of PagerDuty, BlackFog, KPMG, WalkMe, and Cisco 2024 Data Privacy Benchmark Study results, conducted November 2026. Data: data-shared-with-public-ai-tools.csv

Column chart of work information office professionals have shared with public AI tools: any work-related information 88%, emails 43%, meeting notes 40%, customer data 34%, and financial information or confidential documents 31%

Source: PagerDuty Shadow AI Survey, June 2026.

Key Findings

  • Netskope found that generative AI data policy violations doubled over the year, with the average organization logging 223 a month. Source code made up 42% of violations, regulated data 32%, and intellectual property 16%.9
  • Cyberhaven reports that 34.8% of the corporate data employees put into AI tools is sensitive, and 83.8% of enterprise data flowing into AI tools goes to platforms it rates as critical or high risk.11
  • 70% of UpGuard's employees know of sensitive data being shared with AI tools at their workplace, and 23% of CISOs know that passwords and other credentials are being shared. Once that data is in a free tool, the company usually cannot get it back.7,4

IT and Finance Workers Report the Highest Shadow AI Use

Shadow AI is not spread evenly across the workforce, so we compared industry results from TrustedTech's survey of 2,001 US and UK employees with healthcare and seniority data from other studies.

Shadow AI Use and Willingness to Defy a Ban by Industry
IndustryUse unapproved AI toolsWould keep using AI if banned and disciplinedRecognize the risks
IT and telecoms73%48%84%
Finance59%41%80%
HR58%34%n/a
Architecture, engineering, and building54%36%n/a
Manufacturing and utilities52%n/an/a
Healthcare41%n/a76%
Education38%21%76%

Source: AI Services Companies analysis of TrustedTech survey results reported by SecurityBrief UK, conducted November 2026. Data: shadow-ai-use-by-industry.csv

Grouped column chart by industry of employees using unapproved AI tools and those who would keep using AI if banned: IT and telecoms 73% and 48%, finance 59% and 41%, HR 58% and 34%, architecture, engineering, and building 54% and 36%, and education 38% and 21%

Source: TrustedTech, via SecurityBrief UK, September 2026.

Key Findings

  • In US hospitals and health systems, 40% of healthcare professionals have encountered unauthorized AI tools and nearly 20% have used them, and one in 10 has used one for direct patient care (Wolters Kluwer, 518 respondents, December 2025).16
  • Seniority cuts both ways. UpGuard found senior leaders 50% more likely to use shadow AI, and 81% of PagerDuty's office professionals believe leadership plays by different AI rules, rising to 85% at larger organizations. Cyberhaven, which measures all AI use, found mid-level employees using AI tools 3.5 times as much as their managers.7,3,11
  • Knowing the risk does not stop use: in every TrustedTech sector the share who recognize the risks is higher than the share who avoid shadow AI. IT and telecoms workers are also the most likely to say their organization lacks clear AI guidance (57%) or enough training to use AI safely (58%).15

Shadow AI Breaches Cost More Than the Average Breach, and Most Policies Lag

Bans, approved-tool lists, and data loss prevention are how employers try to contain shadow AI, so we put the policy figures next to IBM's breach data to show what happens when those controls fall short.

Shadow AI Policies, Bans, and Breach Costs
MeasureValueSource
Breached organizations with a breach linked to shadow AI20%IBM, 2025
Added breach cost at organizations with high shadow AI useUp to $670,000IBM, 2025
Average cost of a breach involving shadow AI$4.63 millionIBM, 2025
Global average cost of a data breach$4.44 millionIBM, 2025
US average cost of a data breach$10.22 millionIBM, 2025
Shadow AI breaches that compromised customer personal data65% (53% for all breaches)IBM, 2025
Breached organizations with no AI governance policy or one still in development63%IBM, 2025
Organizations with AI governance policies that audit regularly for unsanctioned AI34%IBM, 2025
Organizations that banned GenAI, at least temporarily27%Cisco, 2024
Organizations limiting which GenAI tools employees can use61%Cisco, 2024
Organizations without enforceable data protection policies for GenAI apps50%Netskope, 2026
Security leaders who suspect or have evidence of prohibited public GenAI use69%Gartner, 2025
Knowledge workers who would keep using personal AI tools if banned46%Software AG, 2024
Workers who find workarounds to reach blocked AI apps45%UpGuard, 2025

Source: AI Services Companies analysis of IBM Cost of a Data Breach Report 2025 figures (with VentureBeat coverage), Cisco, Netskope, Gartner, Software AG, and UpGuard, conducted November 2026. Data: shadow-ai-breach-costs.csv

Column chart of average data breach cost in 2025: $4.44 million global average, $4.63 million for breaches involving shadow AI, and $10.22 million US average

Source: IBM, Cost of a Data Breach Report 2025, as reported by VentureBeat.

Key Findings

  • Intellectual property exposed in shadow AI breaches carried the highest cost per record at $178, and 97% of organizations that reported AI-related breaches lacked proper AI access controls.20
  • Penalties are uneven. Of PagerDuty's respondents who used AI that may not have been allowed, 53% received informal feedback to stop and 48% faced formal consequences, and WalkMe found that 78% of executives want to discipline shadow AI use while only 21% of workers have ever been warned about AI policies.3,5
  • Gartner predicts that by 2030 more than 40% of enterprises will have security or compliance incidents linked to shadow AI. In ManageEngine's US and Canada survey, the most common answer from IT decision makers on how to reduce shadow AI, at 63%, was integrating approved AI tools into standard workflows.19,17

What Shadow AI Usage Statistics Mean for Employers in 2026

Unapproved AI use is now an ordinary part of office work. Most surveys from 2025 and 2026 put it between about four in ten and eight in ten workers, and the rates are highest among technical staff, finance teams, and senior leaders. In ManageEngine's survey, the most common unapproved uses were summarizing meeting notes (56%), brainstorming ideas or content (55%), analyzing data or reports (47%), and drafting or editing emails and documents (47%).17

The trend is moving in two directions at once. Companies have moved most AI users onto managed accounts since 2024, but AI use grew so fast that the number of people on personal accounts and the sensitive share of data sent to AI tools both kept rising through 2025. IBM's breach figures show the cost of that gap, and our AI governance and compliance statistics and employee AI training statistics show how far policies, monitoring, and training still trail adoption.

So if you are reporting on shadow AI or setting policy for it, cite the survey's exact question, sample, and date along with the percentage, since surveys of similar office workers range from 31% to 81% depending on how the question is worded and who is asked.

Data Collection and Methodology

We collected published percentages, counts, and dollar figures from 18 studies released between November 2023 and September 2026 . Employee surveys: KPMG and the University of Melbourne (48,000+ people in 47 countries, including 1,019 in the US, November 2024 to January 2025); PagerDuty and Wakefield Research (1,250 office professionals in non-IT roles at organizations with $500 million or more in revenue in the US, UK, Australia, and Japan); BlackFog (2,000 workers at companies with more than 500 employees); WalkMe (3,750 respondents, including 2,050 workers, at enterprises with 1,000 or more employees in 14 countries); UpGuard (1,020 US and UK employees surveyed July 30 to August 11, 2025, and 542 security leaders surveyed August 18 to 31, 2025); Deloitte UK (GenAI Workforce Survey 2026, May to June 2026); Software AG (6,000 knowledge workers in the US, UK, and Germany); Salesforce and YouGov (14,000+ workers in 14 countries, October 18 to 31, 2023); Microsoft and LinkedIn (31,000 people in 31 countries); TrustedTech (2,001 UK and US employees); Wolters Kluwer and CITE Research (518 US healthcare professionals, December 2025); and ManageEngine (700 US and Canadian professionals and IT decision makers).

Telemetry and organization data: Netskope figures come from anonymized Netskope One platform data for October 1, 2024 through October 31, 2025, and its year-earlier values are shown as 2024. Cyberhaven figures come from its May 2024 analysis of 3 million workers (March 2023 and March 2024 data points) and its 2025 AI Adoption and Risk Report. Cisco's 2024 Data Privacy Benchmark Study surveyed 2,600 privacy and security professionals in 12 geographies, and Gartner surveyed 302 cybersecurity leaders from March to May 2025. IBM's 2025 Cost of a Data Breach Report, conducted with the Ponemon Institute, studied 600 breached organizations through 3,470 interviews; the $4.63 million, $4.44 million, and $10.22 million averages are taken from VentureBeat's report on the study, and the 20%, $670,000, 65%, 63%, 97%, and $178 figures from IBM's own summary.

Survey measures differ in wording, sample, and time window, so we show each figure with its own source and do not average them. Where a source gave a phrase such as "more than half" or "about half," we kept the phrase. We calculated products, differences, and ratios from the published values; the estimated 1.8-times growth in personal-account users multiplies Netskope's threefold user growth by the change in personal-account share (47% divided by 78%).

Sources (21)
  1. AI Services Companies analysis of shadow AI survey, telemetry, and breach-cost data from 18 published studies, conducted November 2026.
  2. KPMG, The American Trust in AI Paradox: Adoption Outpaces Governance (April 29, 2025)
  3. PagerDuty, PagerDuty Report Finds Two-Thirds (66%) of Office Professionals Have Used Unauthorized AI Tools at Work (June 11, 2026)
  4. CIO, Roughly half of employees are using unsanctioned AI tools, and enterprise leaders are major culprits, on BlackFog research (January 29, 2026)
  5. WalkMe, State of Digital Adoption 2026 news release (April 9, 2026)
  6. UpGuard, The State of Shadow AI report page
  7. UpGuard, State of Shadow AI news release and methodology (November 10, 2025)
  8. Deloitte UK, What we do in the AI shadows, GenAI Workforce Survey 2026
  9. Netskope Threat Labs, Cloud and Threat Report: 2026 (January 5, 2026)
  10. Cyberhaven, Surge in Shadow AI Accounts Poses Fresh Risks to Corporate Data (May 21, 2024)
  11. Cyberhaven Labs, 2025 AI Adoption and Risk Report
  12. Salesforce, More than Half of Generative AI Adopters Use Unapproved Tools at Work (November 15, 2023)
  13. Microsoft, Microsoft and LinkedIn release the 2024 Work Trend Index (May 8, 2024)
  14. eWeek, Software AG Study Shows Unauthorized Use of AI Tools at Work Rapidly Rising To Over 50% (November 28, 2024)
  15. SecurityBrief UK, IT and telecoms workers most likely to use shadow AI, on TrustedTech research (September 8, 2026)
  16. Wolters Kluwer, Survey finds broad presence of unsanctioned AI tools in hospitals and health systems (January 22, 2026)
  17. ManageEngine, The Shadow AI Surge in Enterprises: Insights from the US and Canada
  18. Cisco, More than 1 in 4 Organizations Banned Use of GenAI Over Privacy and Data Security Risks (January 25, 2024)
  19. Gartner, Gartner Identifies Critical GenAI Blind Spots That CIOs Must Urgently Address (November 19, 2025)
  20. IBM, What data leaders need to know from the Cost of a Data Breach Report 2025 (November 12, 2025)
  21. VentureBeat, Shadow AI adds $670K to breach costs while 97% of enterprises skip basic access controls, IBM reports (July 30, 2025)