Zenity, a security and governance platform for AI agents, has announced an inline prevention solution for Microsoft Foundry, designed to meet enterprise security and compliance needs. This launch coincides with the general availability of Zenity’s inline prevention capabilities for Microsoft Copilot Studio, allowing organizations to implement real-time controls on agents developed, deployed, and operated across the Microsoft agentic stack.
Microsoft Foundry is evolving to provide features that help enterprises build, customize, and deploy advanced agents that orchestrate AI across proprietary and public data, models, workflows, and tools. Zenity’s collaboration with Foundry aims to introduce hard boundaries and runtime enforcement for AI agents. These capabilities are expected to be available soon in preview, developed to align with enterprise standards for security and compliance. Zenity plans to continue enhancing inline prevention for early adopters as new Foundry features are rolled out.
Michael Bargury, CTO and Co-Founder of Zenity, stated, “Securing AI requires understanding of the intent and the full context of the agent. Our work with Microsoft brings agent-centric security directly into Foundry Control Plane, giving enterprises deep visibility and governance, posture management, and real-time detection and inline response when it matters. Together, we’re enabling organizations to implement hard boundaries when adopting AI agents at scale – with the confidence that unsafe actions can be detected, analyzed, and disrupted before they cause harm.”
In addition to the preview capabilities for Microsoft Foundry, Zenity’s integration with Microsoft Copilot Studio is now generally available. Copilot Studio is designed to empower business and IT users to quickly create secure agents for workflow automation. Zenity’s solution offers full lifecycle security, encompassing visibility, governance, vulnerability assessment, and real-time threat detection for these agents.
Zenity’s approach to agentic AI security, which is based on hard boundaries and deterministic runtime controls, includes several key features. These include hard boundaries that provide deterministic, real-time control of AI agent activities at runtime, enforcing strict policy compliance by preventing unsafe behaviors proactively. The solution also offers comprehensive analysis of data flows and tool invocations linked to agents throughout their lifecycle. It is designed to block attacks such as indirect prompt injection and data exfiltration by embedding enforcement directly in the agent execution path, removing ambiguity from agent behavior control. Furthermore, its integration with Microsoft’s agent development ecosystem enables organizations to enforce immutable policy enforcement and hard stops on agent actions.
Sarah Bird, Chief Product Officer of Responsible AI at Microsoft, commented, “Zenity’s integration gives companies real-time control and visibility over the AI agents built with Microsoft Foundry and Copilot Studio. This lets security teams support innovation with the newest AI technologies while keeping data safe and meet compliance needs, helping companies create trustworthy AI agents at scale.”
As companies increase their use of agentic AI through platforms like Microsoft Foundry and Copilot Studio, maintaining security and governance is a priority. Zenity’s runtime enforcement, combined with robust buildtime security posture management, is intended to provide security teams with the necessary tools to ensure agents operate in accordance with company policies, both by design and at runtime, thereby mitigating risk and supporting business innovation.
Zenity’s inline prevention for Microsoft Copilot Studio is generally available, while enhanced capabilities for Microsoft Foundry will be available soon in preview.
