# R-13·cases 0·checked Nov. 15, 2026
Shadow AI and data shared with AI tools
Staff put company or customer data into AI tools the organization hasn't approved or set up, under terms nobody reviewed.
What it is
Data typed or pasted into an AI tool sits with that provider under its retention and training terms, however secure the tool itself is. OWASP's supply chain entry warns that unclear terms and privacy policies can lead to application data being used for training.
NIST's generative AI profile lists data privacy as a core risk, including models leaking or inferring sensitive information about people. In our view, a short policy, an approved tool with reviewed account settings and a clear list of data that must never be entered address most of it.
Documented cases
No case in the register yet. That means we haven't added a verified one, not that it hasn't happened.
Controls to start with
- Review each AI provider's terms and privacy policy, including whether your data trains their models, and re-check when they change.LLM04:2026, mitigation 1
- Send AI tools only what the task needs. Say in policy which data classes must never be entered.LLM02:2026, mitigation 2
- Name the approved tools and accounts, so staff have a sanctioned option instead of a personal one.Editorial guidance
Questions to ask vendors
Our suggestions, not taken from a framework.
- Which AI tools are staff using today, on which accounts?
- For each approved tool: does the provider train on our data, how long is it retained, and can we export or delete it?
Sources
- OWASP Top 10 for LLM Applications 2026. OWASP GenAI Security Project, Aug. 3, 2026.
- NIST AI 600-1, Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile. National Institute of Standards and Technology, July 2024.