A note before the map. Company setups vary enormously, from firms that monitor almost nothing to firms that log almost everything, so this is the general picture rather than a description of your employer. Nothing here is legal advice, and your IT and HR policies are the actual rules where you work.
There is a small flinch that has become part of modern office life: the tab switch when someone walks past, mid-conversation with ChatGPT about the report you are supposed to be writing unaided. Survey research suggests that most professionals who use AI at work do so without their manager's knowledge, while roughly half actively conceal it.
That concealment says more about workplaces in transition than it does about wrongdoing, as the wider question of whether using AI counts as cheating makes clear. It also raises the question almost nobody asks aloud, because asking can sound rather like an admission: what can they actually see?
The honest answer has three zones, and knowing which one you are standing in matters more than any amount of tab discipline.
The short answer. On a company-provided AI account, assume everything is visible, because enterprise tools such as Copilot and ChatGPT's business tiers are built with administration and logging as features. On a personal account used over a work device or network, assume the fact of your visits may be visible even where the content is not, since workplace monitoring of devices and traffic is commonplace. On a personal account, personal device and personal connection, your employer sees nothing from its side, although your obligations still apply to what you paste. The realistic goal is not hiding better. It is needing to hide less.
The three zones
| Where you are | What is realistically visible | What that means |
|---|---|---|
| Company AI account, such as Copilot or a work ChatGPT tier | Potentially everything. Enterprise tools are designed for administration, with logs, usage reporting and retention that your IT team may be able to access | Work in these as if a reasonable colleague could read over your shoulder, because in principle one can |
| Personal account on a work device or network | The fact of use far more readily than the content. Monitored devices and networks can reveal which sites and apps you used and when, even where conversations stay encrypted | The pattern of visits is the exposure, so the more useful question becomes why you feel the use needs hiding at all |
| Personal account, personal device, personal connection | Nothing from your employer's side | Your privacy from the employer is complete here, but your professional obligations travel with you |
The table is deliberately cautious. A work laptop can carry monitoring software that sees more than a network alone, and company configurations vary too widely for anyone outside your IT team to promise exactly what is logged. If you need the precise answer, the policy and the people running the system are more useful than folklore from the office kitchen.
The part that matters more than visibility
The instinct to hide AI use usually comes from a vague guilt about the tool, and that guilt is misplaced. Using AI for drafting and thinking is becoming an ordinary part of how work gets done. The thing that deserves your caution is not being seen with a ChatGPT tab open, but what enters the conversation.
The paste is the real exposure, and it is the same exposure in every zone. Confidential material typed into a personal consumer account from your own sofa is a bigger problem than a harmless brainstorm on a monitored laptop, because everyday AI accounts may retain or use what you type.
If your company provides a sanctioned tool, its occasional dullness is the price of it being the place where sensitive work belongs. The split between Copilot and a personal ChatGPT account is mostly about that containment rather than which logo produces the cleverer paragraph. If no sanctioned tool exists, the discipline of keeping identifying details out of consumer tools is yours to run alone, and the hypothetical method gives you a practical way to keep the useful shape of a situation without exposing the people or company inside it.
Living honestly in the middle zone
Most quiet AI use happens in the middle zone: a personal account on a work machine. The sustainable posture there is simple enough to state, although it requires rather more backbone than a swift tab switch. Use it for work you would defend in a direct conversation, including drafting, structuring, thinking and learning, while keeping confidential material out regardless of who can see.
The strongest form of cover is not concealment but competence, because someone whose work is visibly good has remarkably little explaining to do about their methods. The concealment figures describe a culture in transition rather than a mass of wrongdoing, and transitions end eventually.
The people best placed when this one settles will be those whose habits never needed hiding in the first place. They used the tool where it helped, kept the sensitive material out, checked what travelled under their name and could explain the arrangement without lowering their voice.
Not sure where to start?
Work you'd defend starts with the right tool.
The free finder tells you which AI to open for the task in front of you, in about 60 seconds, and the full “AI, sorted.” reference lands in your inbox.
Try the free finder →One question in, one answer out.
Clair helps non-technical professionals know when to trust their AI, when to check it, and when to skip it.