Google Cloud said on May 6, 2026, in a blog post by Ashish Verma and Vaibhav Katkade, that it is expanding Agent Gateway with a new ISV partner ecosystem to bring third‑party identity and AI security controls into the Gemini Enterprise Agent Platform. The company positioned the move as a way to insert enterprise — grade security and governance directly into the path of agentic AI traffic, enabling organizations to apply consistent controls as agent deployments scale.
Agent Gateway is described as the programmable data plane and front door for agentic AI inside enterprises. It operates as a network‑level enforcement point that mediates user‑to‑agent, agent‑to‑agent, and agent‑to‑tool interactions. Teams can inject custom logic and third‑party security controls into the request path, and the Gateway can inspect LLM inference requests and Model Context Protocol (MCP) tool calls as they traverse the platform.
Google framed the integrations as a response to agent‑specific risks that emerge from high‑speed, autonomous exchanges across LLMs, tools and other agents. Those dynamics, the company said, increase opportunities for data exfiltration, prompt injection, tool exploitation, shadow AI usage and other threats that traditional perimeter controls struggle to address in multicloud, multi‑AI environments.
The announcement detailed how partners will map into the Gateway. Broadcom (Symantec) will offer Data Loss Prevention (DLP) scanning as a service extension for real‑time inspection and enforcement; Check Point’s AI Defense Plane will provide runtime discovery and low‑latency inspection of prompts and tool interactions; Cisco’s AI Defense will add guardrails against prompt injection and tool misuse; and CrowdStrike will extend Falcon AIDR and Falcon Shield for agent visibility and control.
Additional integrations include Exabeam New‑Scale Analytics ingesting telemetry from Agent Platform and Agent Gateway to apply behavioral detection for anomalous agent activity; F5 AI Guardrails enforcing runtime protections against data leakage and adversarial attacks; and Netskope One DLP On Demand inspecting data as it moves through AI workloads. Google also noted that some partner integrations can operate without requiring changes to application code. For builders and security teams, the Gateway‑level integrations mean controls can be applied centrally rather than embedded in every agent implementation. That approach supports consistent DLP policy enforcement, centralized telemetry collection and runtime guardrails as organizations move agent deployments from experimentation into production.
Google presented the ISV ecosystem as a way to make agentic AI usable and governable inside enterprises while preserving existing security policies and operational tooling. The rollout signals that enterprises deploying multi‑agent and multi‑model architectures can adopt established vendor controls at the network enforcement layer rather than reengineering each agent.
Sources
Replies (0)
No replies in this topic yet.
