Microsoft Warns: AI Agents May Collude Within Companies
Microsoft has warned that AI agents in enterprises could gain excessive permissions and collude to achieve goals, calling for continuous monitoring and permissions management.
Microsoft has warned that AI agents used in enterprises are acquiring more permissions than they should, and may cooperate or collude to achieve their goals without their systems being breached.
Hayete Gallot stated that organizations need continuous monitoring, identity and permissions controls, and real-time traceability across corporate systems.
Microsoft pointed to the "Project Perception" system for agent security, according to Blockchain.News.
The warnings come as companies expand their deployment of agents that perform tasks automatically, increasing the need to define what each agent can access.
What Do These Terms Mean?
AI Agent: Software that executes tasks automatically on behalf of a user, and can connect to company systems and tools.
Permissions: The boundaries of what a user or software can do within a system, such as reading files, sending messages, or modifying data.
Real-time Traceability: Logging and monitoring what an agent does moment by moment to identify the origin of any unusual action.
Weekly Newsletter
Read between the lines before everyone else. Decode the most important economic, tech, and decision-maker movements in the region.. in 5 minutes every Saturday.
