Privilege Separation

Privilege separation is a security design principle that splits a system into components with different levels of access, so that a compromise of one part does not automatically hand an attacker control of everything else. In AI systems, it means keeping a model's ability to read data, call tools, or take actions limited to what each specific task requires. For enterprise governance teams, it is a practical control to ask vendors and internal builders about, because an AI agent with broad, unchecked permissions can expose sensitive records or trigger costly actions after a single manipulated input.

1 item