How should AI agents respect company permissions?
An AI agent respects company permissions when its answer depends on who is asking. It uses only what that person can already open at work, in every tool it reads. If an agent runs on one shared key, it does not, and anyone who asks gets everything.
Last updated 11 November 2026 · Mert Uzun, Co-Founder
Why it matters
46%
of CISOs are confident they can centrally control what AI agents access. (Okta, Global CISO Insights 2026)
21%
of CISOs govern AI agent access with shared credentials. (Okta, Global CISO Insights 2026)
90%
of professional developers use AI coding agents every week. (JetBrains, August 2026)
How it works in practice
Agents are now in daily use, and most connect to your tools with one key.
A shared key means an intern and the CFO get the same answer.
Check it with a test. Ask the same question as two people with different access. The answers should differ.
Check what changes when someone is removed from a channel or a drive. Their answers should change too.
Check where a tool has no permissions of its own. The company hierarchy should decide.
How Mithrandir does it
Mirrors each tool's permissions, and follows your hierarchy where a tool has none.
Ethical walls on top, in every agent.
0 leaks measured across 30,000+ entries.
Questions
- Why not give the agent one service account?
- A service account sees everything it was granted, so everyone who asks sees the same.
- Do permissions apply in every agent?
- They should. With one memory behind every agent, the same rules apply in Claude, ChatGPT and Cursor.
- What about legal walls?
- A wall hides walled information from anyone on the wrong side, in every agent.
