LANDMARK · 5 MIN · JUDGE
Privacy before prompting
After this landmark, you can strip a prompt down to the minimum data the task needs and know what happens to what you send.
Pasting into a chat box feels private, but the text leaves your device and lands on someone else’s servers, where it may be stored, reviewed by staff, or used to train future models depending on the product and its settings. That is fine for a lot of work and wrong for some of it. Two moves keep you safe. First, minimise: include only the data the task actually needs, and replace real names, account numbers, health details, and client specifics with placeholders where the task works just as well without them. Second, know the destination: a consumer free tier, an enterprise account with a no-training agreement, and a self-hosted model are three very different places for the same sentence to go. The rule of thumb — don’t paste anything you couldn’t email to a stranger — is crude but catches most mistakes before they happen.
This is the canonical concept. It stays the same across learner lenses so personalization never changes the underlying facts.
◌ EVERYDAY LENS
What this looks like for you
Before pasting, ask “does the model need this exact detail to help me?” Usually a summary or a placeholder works — swap real names and numbers for “a friend” or “$X.” Assume anything you send could be stored, and keep the genuinely private out.
MAKE A DECISION
You want AI help drafting a reply to an angry customer. The original complaint has their full name, order number, and home address. What do you paste?
CARRY THISTake a prompt you’d send today and rewrite it with every unnecessary personal detail removed.