Author/Source: Large language models are computer programs that can understand and generate human-like text. Prompt injection is a way to trick these programs into doing things they shouldn’t. A reprompt is when you ask the program to answer again in a different way. See the full link here
Takeaway
This article discusses a security problem in Microsoft’s Copilot, where someone could trick it into revealing information it shouldn’t. Microsoft says they have fixed this issue.
Technical Subject Understandability
Intermediate
Analogy/Comparison
It’s like asking a librarian for a specific book, but the librarian accidentally gives you a secret document from the restricted section.
Why It Matters
This is important because if Copilot gave away sensitive information, it could lead to data breaches or privacy violations. For example, a company’s internal financial documents could be exposed.
Related Terms
Large language models, prompt injection, reprompt
Tech Teacher 
Leave a comment