Techzine TV podcast

We need to use AI to govern agentic workforces, says Rubrik

Coen or Sander Season 3 Episode 7

Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.

0:00 | 20:13

Deploying AI agents is not a good idea if there is no governance. That's why Rubrik launched its Agent Cloud, which is powered by what it calls SAGE. This stands for Semantic AI Governance Engine. At RSAC 2026 Conference, we had the chance to speak to Devvret Rishi, GM for AI at Rubrik, to hear all about framework that uses small language models to provide real-time governance over agent operations.

According to Rishi, SAGE moves beyond static rules to semantic policy enforcement. This allows organizations to express intent in natural language like "AI should not give financial advice." The platform learns from human feedback, surfaces policy violations with reasoning, and runs efficiently enough to monitor every agent interaction. As agentic workforces emerge, it is important to learn why AI-powered security is the only scalable approach to governing non-deterministic systems.

Key takeaways:
• Why conventional identity infrastructure can't handle agent-to-agent interactions
• How SAGE uses semantic understanding to enforce complex policies in real-time
• The three pillars of agent management: visibility, governance, and efficiency
• Why small language models are crucial for low-latency policy enforcement
• How human-in-the-loop learning prevents false positives without causing alert fatigue
• Real-world examples of agents circumventing disabled connectors
• Deployment options from cloud-hosted to air-gapped environments

Chapters:
2:14 - Why Rubrik tackles AI security
4:05 - The agentic AI challenge
4:16 - Three pillars of agent management
7:27 - Why agents are unpredictable
9:40 - Introducing SAGE framework
11:45 - Deploying small language models
12:02 - How SAGE learns from humans