Google proposes Beyond Zero security paradigm for AI agents

Google security engineers outlined a new cybersecurity paradigm called Beyond Zero, an evolution of Zero Trust for the AI-agent era. The model rests on five principles: authorization by individual actions and resources, blended static and dynamic policies, automatically enriched context, autonomous investigations triggered by risk signals, and on-demand verification challenges for users and agents.

• Beyond Zero evolves Zero Trust to address risks from autonomous AI agents
• Authorization is evaluated per action and resource, not per application access
• Risk signals autonomously trigger investigations and containment measures
• The concept was presented by Google's Heather Adkins and Archana Ramamoorthy