Microsoft now treats AI agents like new employees nobody fully trusts yet :)
Every agent gets its own ID, starts with zero access, gets only what one task needs and loses it when the task is done. If it drifts away from the task its risk score goes up and it can lose edit rights, need a human to approve the next step, or just get shut down. Their deputy CISO says they already run "hundreds of thousands of agents" inside the company.
Via GeekWire: https://www.geekwire.com/2026/microsoft-unveils-tools-to-monitor-and-corral-rogue-ai-agents/
In summer I wrote that an AI agent without clear rules is just another employee with different habits. Now Microsoft sells the rules as a product 😁
Get an email when Kristjan publishes
More from
Kristjan Saarik →
