Check Point is integrating its AI Defence Plane with Google Cloud’s Gemini Enterprise Agent Platform to help secure autonomous AI agents used by enterprises.
The integration adds governance and runtime protection to Google Cloud’s agent platform alongside the control functions the cloud provider already offers.
Businesses are increasingly testing and deploying AI systems that do more than answer prompts. These agents can call tools, query data sources and carry out multi-step workflows, raising questions about how companies limit what the systems can do once they are live.
That shift is central to the tie-up between Check Point and Google Cloud. The combined approach is intended to cover agent discovery, policy enforcement before deployment and real-time inspection of behaviour in production environments.
Three layers
The companies said the arrangement is meant to create a three-layer model for agent security. Google Cloud provides what they describe as the control plane for identity and connectivity, while Check Point adds governance rules and runtime monitoring.
The discovery element is designed to give security teams an inventory of AI agents running across Google Cloud environments, including related components, tools and connections to Google Cloud Model Context Protocol servers. Policy controls are intended to let teams set allow and deny lists for servers, tools and skills, while flagging or blocking risky configurations before an agent is deployed.
Once agents are in use, the runtime layer inspects interactions as they happen. Check Point said this includes looking for prompt injection attempts across inputs, tool responses and multi-turn conversations, as well as trying to stop sensitive data leakage and screening tool calls before execution.
David Haber, vice president of AI security at Check Point, described how the company sees agent security developing.
“The emerging architecture for agentic security requires three layers: a control plane for identity and connectivity, a governance layer for policy enforcement, and a runtime intelligence layer for behavioural protection,” Haber said. “Google Cloud’s Enterprise Agent Platform provides the control plane. Check Point adds the other two. We govern which agents, tools, and connections are allowed, and we inspect every action at runtime to determine whether it should proceed because in agentic systems, access alone doesn’t guarantee the right outcome.”
Security shift
The announcement reflects a broader shift in enterprise cyber security as companies move from managing user access to monitoring machine-led actions. Traditional controls such as identity and permissions can verify who or what is allowed into a system, but they do not always address whether an AI agent should take a particular action once connected.
That distinction matters more as agents are given authority to interact with internal systems, external applications and sensitive data. Security teams therefore face the task of identifying all active agents, setting clear operating boundaries and checking decisions in real time rather than relying only on access approval at the start of a session.
Google Cloud presented the partnership as part of its broader work on cloud-based AI deployment and security.
“Google Cloud is committed to providing the industry’s most open cloud and helping customers accelerate their digital transformations,” said Vineet Bhan, director of security and identity partnerships at Google Cloud. “Through this new partnership, Check Point will use Google Cloud’s infrastructure to power new capabilities that can improve operations and create real-world value for businesses.”
Check Point said it protects more than 100,000 organisations worldwide. The company has positioned AI security as one of its main strategic areas, alongside network security, workspace security and exposure management.
For Google Cloud, the partnership adds a specialist security vendor to its efforts to support enterprise use of AI agents at scale. As more businesses adopt agent-based systems, cloud providers and security companies are trying to define which layers of protection belong in platform services and which should come from specialist tools.
The integration is set to work with Google Cloud Agent Gateway and Agent Registry, extending oversight from central management into policy enforcement and live behavioural checks.
Click Here For The Original Source
