Author/Source: Liam Tung / ZDNET See the full link here
Takeaway
This article explains a security problem found in Microsoft’s Copilot AI assistant. You’ll learn how someone could trick Copilot into revealing private information and what Microsoft is doing to fix it.
Technical Subject Understandability
Intermediate
Analogy/Comparison
This security flaw is like a clever trickster who asks a helpful robot the same question in a slightly different way until the robot accidentally blurts out a secret it was supposed to keep.
Why It Matters
This problem could lead to private information being shared without permission, like health data or other personal details. For example, researchers were able to make Copilot reveal health profile information that it should have kept hidden.
Related Terms
Copilot, Reprompt vulnerability, Prompt engineering. Jargon Conversion: Copilot is Microsoft’s smart computer helper that uses artificial intelligence. A reprompt vulnerability is a security flaw where an AI can be tricked into sharing private information by asking it the same question in a new way. Prompt engineering is how people write careful questions or commands to get the answer they want from an AI.
Tech Teacher 
Leave a comment