The Leak
Three Samsung engineers pasted secret source code into ChatGPT. They weren't spies. They were trying to ship faster.
25 minutes. Branching decisions, a redaction lab, and a classification sprint. A leak-risk score that follows you.
How this works: you play Sam, an analyst under deadline. Every choice about what goes into a prompt changes your score. Sound on for the full effect.
Three paste jobs. One company-wide ban.
Spring 2023. Samsung lets engineers use ChatGPT. Within weeks, three separate teams paste proprietary source code and internal meeting content into it - one debugging, one optimizing, one turning meeting notes into a presentation.
Nobody was trying to leak anything. They were trying to ship faster. Samsung's response: ban generative AI on internal devices and start building its own internal chatbot.
The most dangerous prompt is a productive one. Chatbot inputs may be stored, reviewed by humans, or used for training. They leave your control the moment you hit enter.
9:41 PM.
Clearwater Health Systems.
You are Sam, a data analyst. A claims report is due at 8 AM and your script keeps throwing an error you can't crack.
YOUR SCREEN
The script pulls from the production database - connection string, credentials, and a sample of real patient records are sitting right there in the code. A public chatbot would spot the bug in seconds.
What goes into the prompt?
The next morning, your manager asks for a summary of a confidential vendor contract.
The chatbot writes great summaries. You have the PDF.
A coworker shrugs: "Everyone uses it. IT knows. It's fine."
Redact before you send.
Tap every segment that must come out before this prompt goes to a public chatbot. Then check your work.
Classify before it moves.
SAFETY
101
This certifies that
has completed Module 1 - The Leak: confidential data and public chatbots on with a final score of /100.
Evidence on record: decision history, lab accuracy, sprint calls, knowledge check. This certificate ID verifies in your firm's admin report. A team certificate issues when every enrolled employee completes all 12 modules.
Your one rule to keep: before anything goes into a prompt - classify it, strip it, or route it to the approved tool. If you'd shout it in the lobby, it's fine. Otherwise, it doesn't go in.