Today, Silverfort announces an integration that brings its runtime Identity Security approach to Agent Gateway, part of Google Cloud Gemini Enterprise, giving teams real-time visibility into AI agent activity and enforcing access decisions as they happen.
After Silverfort was named in Google Cloud’s Next ’26 announcement among the partner services bringing identity and governance administration to Agent Gateway, it is time to announce the new milestone we have been building toward, and it lands at exactly the moment enterprises need it.
A milestone for securing agentic AI
Agent Gateway is a genuinely significant piece of infrastructure. Announced in preview at Google Cloud Next ’26, it provides “an open, extensible, scalable way to enforce centralized governance policies to securely connect agents, models, and tools across runtimes.”
For an Identity Security company, that is exciting news. Every agent interaction can be configured to flow through Agent Gateway, which makes it the natural place to apply Identity Security controls at scale. Silverfort now secures the AI agents running there by connecting directly into the platform’s control plane, so agent owners never have to route traffic through a separate gateway.
Why AI agents change the identity problem
Organizations are adopting AI agents faster than most identity and security teams can inventory, monitor, or control them. These agents authenticate into systems, access sensitive data, and take actions across cloud and SaaS environments, often with broad privileges and limited oversight. The result is a rapidly expanding identity attack surface that traditional security tools were not built to address.
Agents also do not fit neatly into traditional security models. Unlike a human user who logs in once and works within predictable patterns, agents can chain actions across services, escalate broadly privileged credentials and tools, and behave in ways that are difficult to predict or trace. That leaves most organizations unable to answer basic questions: how many agents are running, who owns them, what they can access, and which ones are overprivileged.
These risks also arrive from two directions at once. Alongside the exposure created by an organization’s own agents, adversaries are now using AI-powered attack models built to probe, exploit, and move laterally at machine speed. They are not the same problem, but they converge on one control plane: what an identity is permitted to access, and what it is allowed to do. When agents are overprivileged and attackers are autonomous, the window to act is very small.
What Silverfort delivers on Agent Gateway
Visibility into agent activity
Silverfort automatically discovers AI agents operating through Agent Gateway and maps each one back to its human owner for clear accountability. Security teams get a complete view of agent behavior, ownership, permissions, and access paths across Google Cloud and beyond.
Runtime enforcement controls
Silverfort evaluates agent requests against access policies at the Agent Gateway layer before they reach downstream systems, limiting unauthorized privilege escalation and risky agent overreach. If an agent attempts to access a dataset, SaaS application, or internal resource outside its approved permissions, Silverfort evaluates that request in real time and enforces the appropriate policy response. Unauthorized attempts are stopped in line rather than detected after the fact.
Risk-based prioritization
Not all agents carry the same level of risk. Silverfort analyzes agent exposure based on privilege level, access patterns, and connections to sensitive resources. Overprivileged agents, dormant credentials, and high-risk access paths are flagged and prioritized so teams can focus remediation where it matters most.
Granular access policies and detailed audit trails
Policies are defined by agent identity, target application, and action type, and are managed from a centralized dashboard. Access requests are allowed or blocked in real time, and all agent activity is auditable and tied to enterprise identity governance frameworks and to the human using the agent.
Because the controls sit where agent traffic already flows, organizations gain all of this without deploying additional infrastructure or redesigning how their agents are built. Developers keep building. Agents keep running.
Meeting agents where they actually run
AI agents do not all run the same way, so Silverfort delivers runtime enforcement in two complementary ways. The Silverfort MCP Gateway is a universal inline enforcement layer between agents and the Model Context Protocol (MCP) servers they access, built for custom and open-framework agents operating outside closed, vendor-specific ecosystems.
By connecting directly into the control plane of leading agentic platforms, Silverfort enforces policy without requiring agent owners to route traffic through a separate gateway. That’s the right fit for cloud-managed enterprise environments where scale, platform resilience, and zero workflow disruption are critical.
An expanded collaboration with Google Cloud
The Agent Gateway integration is not Silverfort’s only collaboration with Google Cloud. In June, Silverfort launched an integration with Google Security Operations that pulls live identity risk into SecOps playbooks and pushes enforcement back the other way, letting analysts raise a user’s risk indicator or tighten a service account policy from inside a response case. It ships through the Google Cloud partner integrations catalog.
Both integrations point in the same direction: identity telemetry and runtime enforcement reaching Google Cloud teams inside the tools they already run.
One platform for human, machine, and AI agent identities
AI agents do not operate in isolation. They share access paths with human users, service accounts, and other non-human identities, and they often create overlapping risk across all of them. A compromised or overprivileged agent can become a path to lateral movement, privilege escalation, or data exposure, and protecting the agent alone without visibility into the identities around it leaves gaps that attackers will find.
Silverfort brings AI agents into the same Identity Security Platform used to protect human and non-human identities across cloud and on-premises environments. Because those controls apply at runtime, they hold against both pressures – an agent drifting outside its approved scope, and an attacker moving at machine speed with credentials it should never have been able to use. Enterprises rarely run a single agentic AI platform, and enforcement extends across the wider ecosystem regardless of where agents run.
Availability
The Silverfort integration with Google Cloud’s Agent Gateway is now live. Organizations running AI agents on Google Cloud can get started without installing additional software or making code changes to existing workflows. Visibility and runtime protection begin on day one.
This is the beginning
Agentic AI is moving from experimentation to production faster than any enterprise technology in recent memory, and identity is where its security will be decided. Bringing runtime Identity Security to Agent Gateway is a significant step, and one of many we intend to take as the agentic ecosystem grows.
If you’re new to Silverfort, see how we secure AI agents across every platform and request a demo today. For current customers, click here to get started with the integration now.
