New research from CultureAI has revealed that 81% of workplace AI interactions take place outside enterprise accounts, raising concerns about organisations’ ability to monitor AI usage and protect sensitive information.
The AI security and governance company’s latest report, The Unmanaged Majority: What 1.7 million AI interactions reveal about enterprise control, analysed nearly 1.7 million AI interactions across 40 organisations over six months spanning Q2 and Q3 2026.
The findings highlight a significant gap between how organisations believe employees are using AI and what is happening in practice, with sensitive information detected in approximately one in every seven prompts.
AI Adoption Extends Far Beyond Approved Tools
CultureAI identified 1,064 distinct AI applications across the organisations studied, with an average of 113 applications used per organisation. Nearly half of these applications appeared in just one organisation, while general-purpose chatbots accounted for only 6% of the total.
Despite growing investment in enterprise AI licences and governance policies, just 19% of interactions took place through enterprise accounts. Almost half (47%) involved personal, free or unauthenticated accounts, while a further 34% took place through applications where the licence type could not be identified.
The findings suggest that organisations relying solely on approved AI tools and enterprise licences may have limited visibility into how employees are using AI across their businesses.
The research also found that blocking AI applications could have unintended consequences. In organisations that introduced restrictions, personal account usage increased by approximately 50%, suggesting employees were continuing to access AI tools through less visible channels.
Sensitive Data Exposure Becomes Widespread
The report identified 252,480 sensitive data detections across 100 different categories, highlighting the scale and variety of information being shared with AI applications.
HR records and correspondence were detected in 90% of organisations, followed by company strategy information (85%), health conditions (82%), and legal case details and financial information (72%).
Notably, credentials accounted for less than 1% of all sensitive data detections. Instead, disclosures involving company strategy, finance, HR, legal and health information outnumbered credential exposures by almost 30 to one.
This presents a challenge for traditional data loss prevention (DLP) tools, which are typically designed to identify structured information such as passwords, account numbers and other recognisable data patterns.
More contextual information, such as employee grievances, medical conditions, recruitment decisions and confidential business strategies, can be considerably harder to identify and protect.
Growing Compliance Concerns
The findings also raise questions about organisations’ ability to meet their regulatory and compliance obligations as workplace AI adoption accelerates.
Under UK and EU GDPR, health information may qualify as special category data, requiring additional safeguards when processed. Meanwhile, certain high-risk AI systems used in recruitment and employment decisions will face requirements under the EU AI Act, including human oversight and record-keeping obligations, from December 2027.
For legal professionals, sharing confidential or privileged information with external AI services can also create risks around client confidentiality and legal privilege.
The report suggests that simply approving an AI platform or purchasing an enterprise licence is insufficient to address these challenges. Organisations must also consider what information employees are sharing and whether appropriate safeguards are in place.
Visibility, Not Blocking, Is Key
Commenting on the findings, Oliver Simonnet, Lead Cybersecurity Researcher at CultureAI, said: “AI is now firmly embedded in everyday working life, but our research shows a significant disconnect between how organisations believe AI is being used and what is actually happening. With 81% of AI interactions taking place outside enterprise accounts, it’s clear that traditional approaches to governance are failing to keep pace with employee behaviour.”
Simonnet warned that organisations should not assume that introducing AI policies or approving specific applications automatically gives them control over employee activity.
“Organisations cannot assume that approving an AI tool or introducing a policy automatically means they have control over how it is used. They need visibility into the applications employees are accessing, the accounts they are using and, crucially, the information being shared.”
He added that restricting access to AI tools is unlikely to solve the problem, particularly when employees can turn to personal accounts to continue using them.
“Restricting AI isn’t the answer. Where organisations blocked AI tools, personal account use rose by around half. The better route is to make secure use the easiest option. By continuously monitoring AI activity and applying appropriate safeguards, businesses can protect sensitive information without sacrificing the productivity benefits that make AI so valuable in the first place.”
The findings underline the need for organisations to move beyond static approved-tool lists and adopt more continuous approaches to AI discovery, monitoring and data protection.





