Summary
Google released Semantic Governance Policies (SGP) and the SGP engine in Preview on the Gemini Enterprise Agent Platform, letting teams write plain-English business rules and guardrails that intercept an AI agent's proposed tool calls at runtime and verify them against user intent before they execute.
What changed
Google introduced Semantic Governance Policies (SGP) in Preview for the Gemini Enterprise Agent Platform (formerly Vertex AI). SGP adds Natural Language Constraints authored in plain English, Layered Intent Gating that intercepts tool calls at runtime to block unauthorized actions and data exfiltration, and granular scoping across all tools or specific tools and parameters.
Why it matters
As enterprises move agents from pilots to production, the hard problem shifts from capability to control — proving an agent will not take an unauthorized action. Runtime intent-gating on tool calls, authored in natural language, targets exactly the governance gap that keeps regulated buyers from granting agents real permissions, and does it without redeploying agent code.
Evidence excerpt
Semantic Governance Policies (SGP) and the SGP engine are now available in Preview, providing an intelligent security and compliance layer that evaluates an AI agent's proposed tool calls against user intent and organizational business rules at runtime. Layered Intent Gating intercepts agent tool calls at runtime to verify alignment with trusted user intent and prevent unauthorized actions, rogue tool use, and data exfiltration.