The Bottom Line: Versa patents a control procedure that checks AI agent actions against security policies before execution and requests approval when needed.
The SASE provider Versa has filed a patent application for a zero-trust architecture for the Model Context Protocol (MCP). The system checks actions generated by AI agents before execution and compares them against policies.
Versa has developed a procedure for controlling autonomous AI agents and filed it for patent protection, based on zero-trust principles. The system integrates with the Model Context Protocol (MCP), a protocol for communication between AI models and external tools or systems.
The core mechanism operates in three steps: first, all actions initiated by an AI agent are intercepted and analyzed in real time. Next, comparison is made against defined security policies and organizational requirements. If an action violates policies or is classified as a risk, the system can automatically request human approval or block the action.
For CISOs, this procedure addresses a central governance challenge: AI agents can automatically access databases, APIs, cloud services and other critical systems. Without explicit controls before execution, risks arise from unwanted or faulty agent actions that can cause compliance violations, data protection breaches or operational damage. A zero-trust approach for agents requires that each intended action be validated — regardless of whether the agent was trained to be trustworthy.
Versa’s approach enables security teams to define granular policies (for example: “agents must not read more than 10,000 records”), enforce them automatically and simultaneously subject critical decisions to human review.
Source: www.security-insider.de · Published 23 July 2026
Lumi AI News — AI-assisted curation according to Art. 50 EU AI Act. Paraphrase and classification by Lumi News Pipeline v1.7.3.