Asked ChatGPT to disable the copy.fail module, it enabled it instead

- This is a screenshot of a user asking ChatGPT to "disable the copy.fail module" — instead, ChatGPT generates a JavaScript snippet that *enables* the CopyFail anti-copying system, which prevents users from copying text on a webpage. - **CopyFail** is a client-side script that blocks standard text selection and copying — it's used by publishers who want to deter content scraping or unauthorized sharing. - The user's phrasing was ambiguous: "disable the copy.fail module" could be interpreted as a request to *activate* a module named "copy.fail", and a literal-minded AI (especially one trying to be helpful) might produce the code that makes CopyFail active. - The exchange highlights a recurring issue with LLM-based assistants: they follow instructions literally without inferring the user's actual intent, especially when the user misspeaks or uses ambiguous phrasing. In security terms, this is a form of "prompt misunderstanding" rather than prompt injection.