Back to Blog
    Breach AnalysisShadow AI

    Samsung's ChatGPT Leak: The Original Shadow AI Breach

    5 min readAndeavour Research
    12✓45ATTACK CHAINBroken by AgentGuard
    3
    leak incidents reported within weeks

    Key takeaways

    • →Samsung engineers reportedly entered proprietary source code and internal meeting content into ChatGPT on several occasions.
    • →Samsung responded by restricting generative AI tools company-wide, which hurt productivity and pushed usage further into the shadows.
    • →AgentGuard discovers shadow AI usage and blocks sensitive data flows to unsanctioned models, so you can enable AI safely instead of banning it.
    Attack chain, and where it breaks
    1. 1

      Employees adopt public chatbot

    2. 2

      No inventory of AI usage

    3. 3

      Source code pasted for debugging

      AgentGuard detects shadow AI and blocks classified data from reaching unsanctioned models

    4. 4

      Data retained by external provider

    5. 5

      Company-wide AI ban

    Affected
    Samsung Electronics (semiconductor division)
    When
    Spring 2023
    Impact
    Confidential code and meeting data sent to an external AI service

    What happened

    In 2023, multiple outlets including Bloomberg reported that Samsung engineers had pasted sensitive source code and internal meeting information into ChatGPT. Samsung then restricted employee use of generative AI tools on company devices.

    Why bans fail

    Bans don't remove demand. They push AI use onto personal devices and unmanaged accounts, where you have even less visibility. Today the same pattern repeats with shadow agents: autonomous tools employees connect to corporate data without review.

    How AgentGuard would have stopped it

    AgentGuard discovers shadow AI and shadow agents across your environment, classifies the data flowing to them and enforces policy in real time. Engineers keep their productivity, and proprietary code never reaches an unsanctioned model.

    The lesson

    Every incident in this series shares one root cause: an AI system acted with real authority, and nothing in the stack was checking whether that action made sense for its task, its Principal or its environment. That's the gap AgentGuard was built to close, with zero integration and deployment measured in days.

    See AgentGuard on your own environment

    Zero integration. Live in days. We'll show you what your current stack is missing.