AI at work · Transparency
Do you have to tell your employer you use ChatGPT?
Data-loss firm Cyberhaven's 2026 report estimated that about 4 in 10 workplace AI interactions carry sensitive data — which is why the question « do I tell my manager? » matters less than « what am I pasting in? ». Here is the real rule, without paranoia and without naivety.
The rule, in three points
- No law makes you announce every personal tool you open. In most countries, using ChatGPT on your own account is not, by itself, something you owe anyone a declaration for.
- Your duty of loyalty already exists: your contract, confidentiality, trade secrets. If your company has a written AI policy, it applies. Reading it takes ten minutes and beats discovering it in a meeting.
- If ChatGPT is explicitly banned where you work, do not route around it: that is the one case where the risk becomes disciplinary, and a personal account protects you from nothing.
What is fine, what never is
| Situation | Verdict | Why |
|---|---|---|
| Rephrasing a paragraph you wrote yourself | Fine | Your text, your level of confidentiality. Nothing identifiable leaves. |
| Asking for an outline, a template, a document structure | Fine | The request contains no company data. |
| Pasting a client email with names and company | Never without approval | Personal data and confidentiality: the risk is not the tool, it is your contract. |
| Pasting a spreadsheet formula with the company's real figures | Never without approval | The numbers are not yours, even half-anonymised. |
| Pasting source code or a contract | Never | The heart of trade secrets. No rewrite walks that back. |
| Asking how to say no to your manager's request | Fine | Anonymise the situation and you are done. |
| Pasting a colleague's payslip or HR document | Never | A third party's personal data. Full stop. |
Telling your manager: one sentence is enough
You do not need a confession or a dossier: one sentence in a one-to-one or in writing does the job. "I use ChatGPT on my own account to save time on formatting and templates — never with client data. If you'd rather set the rule, tell me and I'll follow it." That message does three useful things: it makes the use visible, it shows you know where the line is, and it puts the decision where it belongs — the rule is your employer's to make. Most of the time the answer is "fine, just no confidential documents", which is exactly the line in the table above.
The line that actually matters
The question "should I tell?" hides the real red line: what goes into the tool. A declared use that pastes whole contracts is riskier than a quiet use that pastes nothing but templates. Every skill in the Eternal Intern pack is written with that line in mind: it states what you can hand over, what you never paste, and nothing in it is designed to send anything automatically.
LUCIDIA, the company behind this site, is itself run end to end by AI agents on NanoCorp.
Use ChatGPT with no grey zones, job by job
The Eternal Intern pack installs eight skills and a prompt library in your own ChatGPT or Claude subscription, with the confidentiality rules written into every skill: what you can hand over, what you never paste. You answer six questions, the ZIP arrives by email. €19.99, one-time. The AI subscription is not included — the pack makes yours useful, and nothing goes out without your read-through.