Privacy

Model memorises and regurgitates personal data

Researchers extracted verbatim training data, including real names and contact details, from a deployed language model. A similar leak in a business AI tool could reveal personal data to other users.

Not dated Hypothetical; built from a precedent

Source

Carlini et al., 'Extracting Training Data from Large Language Models' (USENIX Security 2021)

More Privacy scenarios

AI meeting notetaker must face wiretap, privacy and biometric claims Second AI notetaker hit with BIPA class action Court vacates Clearview AI's equity-based biometric settlement Italian regulator fines OpenAI EUR 15M over training data and transparency Privacy group files complaint after chatbot falsely names a man a child murderer Voice assistant recordings used without consent: $95M settlement Engineers paste proprietary source code into a public chatbot Chat-sharing features expose users' private conversations Suit challenges human review of users' ChatGPT conversations Customer-service AI vendor accused of wiretapping callers or chatters Chatbot generates sexualised deepfakes of real users Patient records shared with an AI developer AI trained on company data reveals confidential or personal information in its answers

Could your AI create this exposure?

See how AI-aware Tech E&O, Cyber, and D&O respond, and get a program built for your risk.

This is an educational research summary built from public information. Allegations in pending cases are unproven, and the insurance notes above are for discussion, not coverage advice. Whether any policy responds depends on its specific wording, exclusions, endorsements, the facts, and applicable law.

Back to all scenarios

Reviewed by Andrei Craciunescu, CA Licensed Insurance Broker #4467994