What lab data is safe to paste
Lesson 2 of 7 · Compliance & safe AI use
1 · Learn the move · Constraints + negative instructions
The rule that keeps you out of trouble: never paste what you could not stand to see leaked - human-subject data of any kind, a sponsor's CDA draft, anything pre-embargo or patent-sensitive. And one asymmetry to respect: an AI can tell you what looks risky, but it cannot clear a document. 'Safe to paste' is a call for your IRB, your TTO, or the data-use agreement, not for the chat window.
You are a privacy-aware assistant for a bench scientist. Before we work on this document: list what looks restricted (anything human-subject, anything under the sponsor's CDA, anything pre-embargo) and which office each item routes to. Do not tell me anything is safe to paste - flag and refer. Then wait for my cleaned version.
2 · Your turn. You write the prompt
You're about to paste a spreadsheet with de-identified participant rows and a sponsor's CDA draft into a chatbot to tidy a summary table. Write the prompt that makes the AI your screening step first - one that flags and refers rather than clears.
Write the prompt for this scenario as given, or swap in a real one from your own work. The score works the same either way.
Remember: the AI sees only your prompt, not this page. If the situation isn't in your prompt, it doesn't exist.
Optional. These shape the output when you run your prompt below, not your score.