The Unpredictability Problem
The rapid adoption of AI agents and applications presents a unique challenge for enterprise security teams. Unlike traditional deterministic software, AI systems operate on probabilities, not fixed rules. Niv Braun, CEO of Noma Security, explains that this fundamental unpredictability creates a perfect storm where the potential blast radius is enormous, yet time to market pressures leave security teams scrambling to keep up.
A Unified, Context Aware Strategy
Braun argues that effective AI security requires a holistic framework built on two core principles: flexibility to absorb fast evolving technologies like the Model Context Protocol (MCP), and deep contextualization. Security teams must connect posture management, access controls, and runtime monitoring into a unified signal. Without visibility into runtime behavior, it is impossible to make informed recommendations about configuration and access permissions for an AI agent. This approach shifts security from reactive rule enforcement to proactive, context aware governance.
Identifying Legitimate vs. Malicious Actions
A key practical challenge for security platforms is distinguishing legitimate agent actions from real threats. Braun notes that early partnerships between AI providers and security vendors can help enable secure by design capabilities. A unified security platform is also vital, as it outperforms siloed point products by correlating signals across the entire AI stack. As AI continues to evolve, the ability to understand and control an agent’s context will determine whether enterprises can safely harness its power.
Source: Healthcareinfosecurity
