AI is changing what access means.
Give someone an AI assistant and the permissions on their account can stay exactly the same. What they can do with those permissions can change overnight.
The same account can reach further
Consider a project folder that a former team member can still read. Yesterday, finding a sensitive detail meant opening dozens of documents. Today, an assistant with that same access can pull the detail into a summary in seconds. If someone shares the summary, the information travels beyond the folder that governed it.
Agents and service accounts introduce new identities to govern. A change just as consequential is happening inside the accounts people already use. AI is finding its way into mail, search, spreadsheets and code editors. Each integration changes what someone can discover, combine or act on. As those capabilities spread, the risks of a stale permission or an overly broad role grow just as quickly.
Why we built this with AI
AI generated nearly 100% of this site and the Identitybox tool. We chose to build this way to explore a practical question. Can the technology accelerating these changes also help people understand and govern access?
That question shapes how the tool works. A reviewer can use an authorized assistant to ask why someone who moved out of Finance still has ledger access. The assistant can look up the account and its role assignments. If it also has access to the transfer record, it can bring that evidence into the review. The finance owner can then judge whether the person still needs access, with the relevant records in front of them.
As AI becomes part of everyday work, “Who has access?” deserves a second question. What does that access let them do today?