Noma Security, the unified AI security and governance platform, has officially announced the launch of its Agentic Risk Map (ARM), which happens to be the industry’s first visualization technology, geared towards mapping the blast radius of autonomous AI agents.
According to certain reports, this particular development arrives bearing an ability to secure agentic AI across discovery, posture management, and runtime protection.
To understand the significance of such a development, we must take into account how AI agents generally operate autonomously across digital ecosystems. This operation involves accessing databases, executing code, sending communications, and making decisions that ripple across enterprise systems. You see, leveraging MCP (Model Context Protocol) servers, these agents can seamlessly connect to an expanding universe of third-party tools and services, thus significantly increasing their reach and potential impact.
Such a complex system, like you can guess, makes it impossible for security teams to visualize and understand the full blast radius of a compromised agent.
Against that, Noma’s latest brainchild brings forth three integrated phases i.e. discovering all agents across the enterprise (including shadow AI), assessing and managing their security posture before and during deployment, as well as protecting them at runtime with continuous monitoring and immediate containment capabilities.
Talk about these phases on a slightly deeper level, we begin from Discovery, which focuses on facilitating complete visibility into the Agentic attack surface. This translates to how Noma automatically identifies and catalogs all AI agents across the enterprise, including shadow AI and unauthorized deployments.
More on the same would reveal how it also discovers every MCP server, toolset, API connection, and agent-to-agent relationship to reach upon a complete inventory of the agentic attack surface. In case that wasn’t enough, organizations can also gain unprecedented clarity into what agents exist and where they’re deployed.
Next up, there is the step of proactive risk management. Here, Noma’s Agentic Risk Map will flip the invisible maze of agentic infrastructure for actionable intelligence through comprehensive visual maps of an organization’s entire agentic ecosystem.
Alongside that, having ARM in the mix also reveals the true scope of risk by mapping agent-to-agent (A2A) connections, tool and MCP server access, cross-system dependencies, and permission chains. The overarching idea here is to expose the cascading pathways through which a single compromised agent could trigger unauthorized money transfers, exfiltrate data, or move laterally across the organization.
Team can even assess potential blast radius before deployment through a closer look at how new agents will connect to existing infrastructure, while simultaneously performing red team testing on agents before deployment. The underlying ARM also enables security architects to create blueprints for scoping permissions and implementing controls based on mapped relationships, preventing excessive agency.
Another detail worth a mention is rooted in Agent Runtime Protection. In essence, Noma continuously monitors agent behavior against the established baseline to detect anomalous actions, such as unexpected tool invocations, unauthorized agent-to-agent communications, suspicious cross-system access patterns, or potential prompt injection attacks. The platform is markedly well-equipped, at launch, to provide immediate containment capabilities to stop surging damage before it spreads across enterprise systems.
Among other things, it ought to be acknowledged that Noma’s platform is capable of supporting full spectrum made up with AI agent platforms and infrastructure. These platforms are presently understood to include Microsoft Copilot Studio, ServiceNow, Salesforce Agentforce, Google Agentspace, Azure AI Foundry, Google Vertex AI, AWS Bedrock AgentCore, LangChain, CrewAI, Google ADK, and OpenAI SDK.
On top of it, they can even support several coding agents, such as Cursor, GitHub Copilot, and other AI-powered development and productivity tools.
“Security teams are flying blind when it comes to AI agent risks,” said Niv Braun, CEO and Co-founder at Noma Security. “These agents don’t just touch one system, they span departments, tools, and workflows. A seemingly harmless Customer Support Agent, if compromised, can cascade into unauthorized money transfers, sensitive data exfiltration, and malicious emails sent to customers or employees for lateral movement. Organizations need more than point solutions. They need complete visibility, proactive risk management, and runtime protection working together. That’s what we’ve built.”