Safe everyday use
Before anyone practices real workflows, they need a few habits that prevent the most common harms. Those harms are exposing private information, passing along a confident error, or falling for an urgent scam. These habits are simple, and they are the floor everything else stands on.
By the end, you will be able to
- Decide what information is safe to put into an AI tool, and what is not.
- Understand why even free or approved tools still need care.
- Verify an answer before you use it.
- Recognize a warning sign, stop, and report it the right way.
Lesson 1
A prompt is a disclosure
Whatever you type into a tool, you are disclosing. The first habit is deciding what is safe to share.
Every time you paste something into an AI tool, treat it as sharing that information. So the question before you type is simple: what kind of information is this, and is it safe to share here? A simple way to sort it, and your organization may use its own labels, is by how sensitive the information is.
- Public: already published, safe to use.
- Internal: fine within the organization, and usually fine in an approved tool.
- Confidential: private plans or undecided matters, keep out.
- Restricted: the most tightly controlled information, like Social Security numbers, health records, or immigration status, keep out.
- Beneficiary-sensitive: any other detail about the people you serve that could harm them if exposed, keep out.
Important
When you need to work with real material, remove the private parts first, or use made-up stand-in information. Aggregate numbers are usually fine; a named person's details are not.
Try it
Picture something you might paste into an AI tool this week. Which of the five types is it? If it is confidential, restricted, or beneficiary-sensitive, what would you remove first?
RememberTyping is sharing. Sort the information first, and keep private details out.
Lesson 2
Approved tools and accounts
Free, personal, or built-in tools are not automatically safe. Where the information goes matters as much as what it is.
A tool being free, popular, or already inside your software does not tell you how it handles your data. Some tools keep and reuse what you type. Some send it to other companies. Some are personal accounts that your organization has no agreement with.
Try it
Do you know which AI tools your organization has actually approved, and for what? If not, that is a useful question to bring back to your team.
RememberApproved means reviewed for this kind of data. If it has not been reviewed, treat it as public.
Lesson 3
Verify before you use
Never pass an AI answer straight into real work. A short, repeatable check catches the predictable errors.
The habit that protects you most is a quick verification pass before an answer becomes real work. It takes a minute and it catches the failures from the earlier modules.
- Math: recompute any number yourself.
- Sources: confirm any fact or citation against the real source. Assume citations, links, and quotes can be fabricated even when they look real, and open them to check.
- Scope: check it did not overreach or invent a requirement.
- Supported facts: make sure every claim is backed by something you gave it.
- Owner: confirm the right person reviews and approves before it goes out.
A draft says 410 of 500 people is 92 percent. Recompute it: 410 divided by 500 is 82 percent. The tool produced a confident, wrong number. Thirty seconds of checking caught it before a leader ever saw it.
Try it
Next time you get an AI answer with a number in it, recompute the number before anything else. Make that your reflex.
RememberRun five checks before use: math, sources, scope, supported facts, and the human owner.
Lesson 4
Stop, contain, and report
When something feels wrong, the safe move is to stop, verify another way, and report through a trusted route. Urgency is the warning sign, not the reason to skip the process.
Some situations are not about a wrong answer. They are about a threat: a suspected data exposure, a harmful output, an impersonation, or a scam. The habit here is a calm sequence you can run under pressure.
- Recognize the warning signs, especially pressure to move fast or go around the normal process.
- Stop, and do not act on the request yet.
- Verify another way, through a person or channel you already trust.
- Report it through your organization's trusted route.
Important
A common trap: an urgent message that looks like it is from a leader, asking you to bypass the normal payment or approval process. Urgency is exactly the sign to slow down and verify through a channel you trust, not to comply.
Try it
Say the four steps in order from memory: recognize, stop, verify, report. That short chain is your safety net when something feels off.
RememberWhen something feels wrong, stop and verify through a trusted route. Urgency is the warning, not the excuse.
What you leave with
Stop-Check-Escalate card and safe-input matrix
A one-page card for the five verification checks and the stop-and-report sequence, plus a simple grid for what information is safe to put where. Kept near your desk, it turns safe use into a habit rather than a memory test.
This is just for you. It saves on this device only, and nothing is scored.