Corporate account or personal account?

The same AI service can be safe on one account and risky on another. An employee signed in to your company's tenant is covered by your contract: retention terms, admin controls and audit logs. The same employee on a personal account is covered by their own consumer terms. A policy that only looks at the domain cannot tell these two apart.

One address, two very different risks

From a network point of view, both sessions look alike. They go to the same address and use the same application. A web filter sees one destination.

For the organization, they could hardly be more different. On the corporate account you chose the terms. You know how long data is kept, whether it trains a model, and who can read the logs. If someone leaves, the account closes and the history stays with you.

On the personal account you chose nothing. The data sits under terms the individual accepted in a hurry. When that person leaves the company, the chat history leaves with them. You have no log, no retention control and no way to ask for anything back.

Why domain rules fail in both directions

Faced with this, most teams reach for the tool they already have, and it gives them two poor options.

If you block the domain, you break the approved tool. The people who rely on it stop working, or they find another way. Security becomes the department that took away the useful thing.

If you allow the domain, the personal account walks through the same door. You have paid for a corporate service and left the side entrance open beside it.

A popular compromise is a written rule: use the company account only. It is worth having, but it depends on people noticing which account they are signed in to, on a browser where both may be open. Anyone who has sent a work message from the wrong profile knows how that goes.

What account-aware policy looks like

The fix is for the policy to know which account is in use, and to treat the two differently. In practice the rules are short.

• Approved corporate account: allow it, check the content and keep a record.
• Personal account on the same service: warn the person, ask for a business reason, or block.
• Sensitive content in either: the detection engine decides, not the domain.

The third rule matters more than it looks. A corporate account does not make every prompt acceptable. A customer list or a contract may still be outside what your policy allows, wherever it is going. The account decides the starting point, and the content decides the outcome.

What the person sees

Good policy shows up in the moment. Someone pastes a customer list into a chatbot and has not noticed they are on a personal account. Instead of a silent failure, they see a short message: this looks like customer data, this is a personal account, please use the company one.

That message does three useful things. It stops the data. It teaches the rule at the point where it matters. It also gives the person an immediate way to finish the task, which is why they were there in the first place.

Compare that with a block page that gives no reason. One leaves people informed and working. The other leaves them looking for a workaround.

Questions to ask about your own setup

A few questions will show where you stand.

• Can you see which AI accounts people use, not only which services?
• Does your policy treat a corporate and a personal account on the same service differently?
• If someone is stopped, do they learn which account they were on and what to do next?
• When a person leaves, what happens to the AI history created for their work?

Where the answers are unclear, start in monitor mode. Find out how much use goes through personal accounts, and in which teams. Often it is a small group with a real need the approved tool does not meet, and that is a conversation worth having before any rule changes.

The point of the distinction

Approving a corporate assistant and restricting the personal one is not a technical detail. It is what makes a clear policy possible. You can say yes to the tool people want, keep your contract's protections, and stop the data that should not leave, all without breaking the service for the people who depend on it.

Read how AI Data Security separates corporate and personal accounts.