Question 1

Why did an AI agent access data it shouldn't have?

Accepted Answer

Usually because access was granted to the tool the agent runs on rather than to the person asking it a question. A shared service account or an over-provisioned connection gives every request from that agent the same broad access, regardless of who is actually behind it or what they are individually entitled to see.

Question 2

How do you stop an AI agent from exposing sensitive data, such as a Medicare number?

Accepted Answer

By carrying the security classification on the data itself rather than on whichever tool happens to be reading it. In CryspIQ, sensitive fields are protected by default using Microsoft Entra ID security groups and Microsoft SQL Server Dynamic Data Masking at the column level, so an agent inherits the same restriction any other consumer of that field would, wherever it reads from.

Question 3

Does an AI agent see everything a service account can access?

Accepted Answer

It should not. Authorisation should be computed from the real, individual user's identity behind a request, not from a shared bot or service account. An agent acting on behalf of someone should see exactly what that person is permitted to see through any other channel — no more.

One post tagged with "Data Security"

Does Your AI Agent Know What It's Allowed to See?