Safe AI use
Shadow AI without shame.
When people use AI quietly at work, the answer is not panic. The answer is clearer boundaries, safer examples, and better conversations about approved use.
Why this matters
Hidden AI use is often a rules problem, not a character flaw.
Workers may use AI quietly because they are curious, under pressure, embarrassed, or unsupported. Shame does not make that safer. Practical boundaries do.
The shadow AI safety reset.
1
Name the pressure
Are you using AI because work is unclear, too slow, too repetitive, or socially awkward to ask about?
2
Separate safe from risky
Practise with invented examples, public information, and low-risk drafts before using real workplace material.
3
Check the tool
Know whether the tool is approved, personal, monitored, connected to work systems, or outside policy.
4
Ask for boundaries
Turn private guessing into clear questions about tools, data, review, and examples.
Copy-paste prompt: make this AI use safer.
# ROLE
You are my workplace AI safety coach.
# TASK
Help me decide whether this AI use is safe enough to try at work.
# AI USE I AM CONSIDERING
[describe the task]
# TOOL
[approved workplace tool / personal tool / not sure]
# DATA INVOLVED
[public / invented / internal / customer / employee / financial / confidential / not sure]
# OUTPUT FORMAT
1. Risk level: low, medium, or high
2. What data I should remove
3. What tool or policy question I should ask
4. A safer version of the task
5. A low-risk practice example
6. What a human must review before use
# RULES
- If the data may be private or confidential, tell me to stop and clarify the rule first.
- Do not make legal or policy claims.
- Keep the tone calm, not frightening.
Use this before you paste.
Green
- Invented examples.
- Public information.
- Personal practice notes.
Yellow
- Internal process notes.
- Drafts with names removed.
- Summaries checked by you.
Red
- Customer or employee data.
- Commercially sensitive details.
- Anything policy-critical.