You already brought AI into your company. Nobody told you what is allowed.
You opened an AI today to sort out something at work. A difficult email, a spreadsheet that would not balance, a text that would not come out. It worked, you delivered it, and you told nobody.
This is not a confession. It is the normal situation, and this text is about what it creates — which is not what most people assume.
Why nobody says anything
Two reasons, and both are reasonable.
The first is looking like you did not do the work. There is a strange feeling hanging around AI at work, as if using it were cheating on a test — and nobody wants to be the first to find out how the boss reacts.
The second is more practical: saying it can mean losing the tool. If the answer is "then don't use it", you go back to spending three hours on what was taking you twenty minutes. Safer not to ask.
The result is that many people use it, almost nobody says so, and the company carries on believing the subject has yet to arrive.
What happens while nobody knows
Here is the part that usually goes unnoticed, and the only one that truly matters.
While nobody knows, you have become the only person deciding three things that should not be yours alone:
What goes in. Every file you paste is a decision about company data — and you make it with no yardstick, mid-task, at eleven at night. It is the subject of the other text here, and at work it weighs more, because the data is not yours.
What comes out. AI is wrong with the same confidence with which it is right. When you hand the result in, it becomes your work — and the check nobody asked for is the only one that exists.
The credit. If nobody knows how it was made, nobody knows what was checked. And the day someone asks "how did you get to this?" is the worst possible day to explain your process for the first time.
Notice that none of the three is about you having done something wrong. They are decisions the company should have made, and that landed in your lap through its own silence.
The mistake does not become the AI's mistake
This is the line worth keeping, and it has appeared here before in another context: responsibility does not shrink as the tool improves.
If the number is wrong in the spreadsheet you handed in, nobody will say "the AI got it wrong". They will say the report was wrong — and the name on the report is yours. The tool signs nothing.
That is not a reason to use it less. It is a reason to check what you deliver as if you had written it by hand — because, for every practical purpose, you did.
How to raise your hand without becoming the test case
The instinct is to confess: "look, I've been using AI". It is the worst route, because it makes you the subject and opens a conversation about your past instead of the company's method.
What works is the opposite: ask what is allowed, do not report what you did.
"Do we have any guidance on what can and cannot go into an AI tool? I'd like to follow the right rule."
Notice what that sentence does. It asks for no permission, confesses nothing, and moves the subject to where it belongs — and whoever asks first does not become the problem case: they become the person who raised the subject before it became a problem.
If the answer is "we have nothing written" — which is the likeliest answer — you already got what you needed: the absence is now on record, and the decision stopped being quietly yours.
And while the rule does not exist, one line covers almost everything: do not use a personal account for company data. That is the difference between a use someone can audit later and one nobody can even reconstruct.
If you are the one who answers for the company
Then the reading is different, and less comfortable: people are already using it. The question stopped being whether the company adopts AI — it already has, scattered, through people who had nobody to ask.
This is not about hunting down who uses it. It is about writing the yardstick those people are asking for without asking: what can go in, before deciding who uses it. And deciding that is the right of whoever answers for the business, not an obligation to become technical.
The test
Think of the last thing you delivered with help from an AI.
If your boss asked today how it was made, would you have an answer ready — or would you have to invent one?
If the latter, the problem is not the tool you used. It is that you are carrying alone a decision that was never only yours.
This is not legal or HR advice: what your contract and your company allow is written down somewhere, and almost never mentions AI yet — which is exactly the subject of this text.