Welcome to the AI Agent & Copilot Podcast, analyzing the latest AI Copilot and agent developments from Microsoft and its partners, delving into customer use cases, and exploring how AI plus the Cloud helps customers reimagine their business. In this episode, Tom Smith speaks with Michael Bargury, Co-Founder and CTO at the AI agent security software firm Zenity.
Highlights
Zenity’s Focus on AI Agent Security (01:15)
Bargury notes that enterprises are adopting AI at the speed of startups, emphasizing the importance of trust in AI agents and AI-driven technology. Zenity’s value proposition is to help enterprises trust their AI agents, ensuring they are secure and reliable for production systems.
Enterprises Using 6-7 Agent Platforms (03:30)
Bargury says large enterprises often build their own AI agent platforms initially but realize the need to adopt off-the-shelf solutions due to the rapid pace of AI innovation. Enterprises are using multiple agent platforms — “at least like six or seven” — making it challenging for security teams to manage. The shift from building internal platforms to using established solutions like Microsoft and OpenAI is driven by the need to keep up with AI’s fast-paced innovation.
AI Agent & Copilot Summit is an AI-first event to define opportunities, impact, and outcomes with Microsoft Copilot and agents. Building on its 2025 success, the 2026 event takes place March 17-19 in San Diego. Get more details.
New Risks: Agents Going Rogue, Insiders (06:23)
With AI in general and AI agents more specifically, there are two categories of risk: traditional security concerns (e.g., Data Loss Prevention, data sovereignty, misconfiguration) and new risks introduced by AI agents (e.g., agents going rogue, insider risks). The new risks involve agents behaving in unpredictable or immoral ways. The traditional security concerns are similar to those in traditional applications but are increased by the rapid pace of AI innovation.
Customers Need AI Agent Security Programs (13:38)
Bargury emphasizes the need for a comprehensive security program focused on agents, similar to an insider risk program but for AI agents. The program should start with visibility, understanding which agents are touching sensitive data and creating feedback loops to manage risky agents. Zenity’s platform helps discover and remediate risky agents, with some issues being automatically fixed without human intervention.
Microsoft Partnership and Copilot Studio Integration (18:53)
At the recent Build 2025 conference, Zenity announced a partnership with Microsoft to integrate Zenity’s security capabilities into Copilot Studio. The partnership aims to align Zenity’s security program with Copilot Studio’s governance capabilities, ensuring a seamless and secure AI agent experience for customers.
Challenges and Opportunities with MCP and Agent-to-Agent Protocols (22:45)
Bargury discusses security concerns with Model Context Protocol and the Agent2Agent Protocol, such as lack of authentication and access control, and the need for improvements. The broader issue is the loss of control and boundaries as agents communicate and share data across different parts of the network. Despite the current challenges, these protocols have the potential to make AI agents more productive and impactful, but they require careful management to mitigate security risks.
Ask Cloud Wars AI Agent about this analysis
