AI Systems Leak Sensitive Data

AI adoption is expanding the ways personal, confidential, and regulated information can be collected, inferred, retained, exposed, or reused.

The most persistent risks involve employees submitting data to unapproved tools, AI systems and agents accessing internal repositories, model memorization and extraction, and unclear vendor practices around training and deletion. Regulators, courts, and organizations are responding through privacy rules, contractual controls, access restrictions, auditing, and privacy-enhancing technologies, but protections remain uneven across jurisdictions and use cases.