Four in five workplace AI interactions happen outside enterprise accounts, according to new research from CultureAI. The AI security and governance firm analysed 1,668,999 prompts and 161,599 file uploads from 40 organisations over six months from April to September 2026, before its own controls were applied.
Only 19% of interactions ran through enterprise accounts. Of the remaining 81%, 47% used personal, free or unauthenticated access and 34% had no identifiable licence type, mostly on applications that offer personal accounts only. CultureAI observed 1,064 distinct AI applications, around 113 in the average organisation, and general-purpose chatbots made up only about 6% of them.
The study recorded 252,480 sensitive data detections across 100 data types, roughly one for every seven prompts. HR information was disclosed in 90% of organisations, company strategy in 85%, health information in 82%, and legal and financial information in 72%. Credentials accounted for under 1% of detections, so company strategy, finance, HR, legal and health disclosures outnumbered them by almost 30 to one.
Blocking did not appear to help. In organisations that blocked AI applications, personal account use rose by around 50% afterwards. CultureAI argues that pattern-based tools built to catch passwords and account numbers will miss most of what staff actually share, because a grievance, a diagnosis or a board forecast has no fixed format.
The report points to compliance exposure. Health data can be special category data under UK and EU GDPR, and from 2 December 2027 high-risk AI systems used in recruitment and employment decisions fall under EU AI Act requirements including human oversight and record-keeping. The data comes from organisations that already use the CultureAI platform, which the authors note when describing the sample.
To stay across the latest in cloud, AI and enterprise tech analysis from Compare the Cloud, subscribe to our weekly newsletter at https://www.comparethecloud.net/newsletter