Key Takeaways
- AI agents and non‑human identities (NHIs) now outnumber humans by roughly 100 : 1, creating a major blind spot in enterprise security perimeters.
- Cisco’s acquisition of Astrix Security adds capabilities for discovering, governing, and protecting agentic and NHI activity across the full lifecycle—from provisioning to decommissioning.
- Astrix’s platform addresses three core pillars: discovery & governance, access & lifecycle management, and threat detection & response.
- The move follows Cisco’s recent purchase of AI observability firm Galileo Technologies, signaling a broader strategy to embed AI‑agent security into its Splunk observability and networking portfolio.
- Industry data shows only about a quarter of organizations can currently control agent actions with proper guardrails, highlighting a growing gap that Cisco aims to close.
Discovery and Governance for AI Agents
“Discovery and governance for AI agents: Provide a map of the org’s agentic activity, vet policy to resolve hygiene issues, reduce attack surfaces, and prevent compliance violations.” This opening line from Astrix’s product description captures the first challenge security teams face: simply knowing what agents exist and what they are allowed to do. In many enterprises, agents spin up in development sandboxes, CI/CD pipelines, or as part of third‑party integrations, often without a central registry. Astrix’s discovery engine continuously scans cloud workloads, container platforms, and identity providers to build a real‑time inventory of every AI‑driven service account, API token, or robotic process. Once identified, the platform evaluates each agent against organizational policies—checking for excessive privileges, outdated credentials, or misaligned scopes—and flags hygiene issues that could be exploited. By enforcing least‑privilege guardrails and continuously validating policy adherence, Astrix shrinks the attack surface and helps organizations stay ahead of regulatory requirements such as GDPR, HIPAA, or emerging AI‑specific statutes.
Agentic Access and Lifecycle Management
“Agentic access and lifecycle management: Manage AI agents and their NHIs from provisioning to decommissioning.” Effective security does not end at discovery; it must follow agents through their entire existence. Astrix treats each non‑human identity as a first‑class citizen in an identity governance framework, applying the same rigor used for human users. When a new agent is provisioned—whether via a DevOps pipeline, a low‑code automation tool, or a machine‑learning model deployment—the platform automatically assigns it a role based on predefined templates, injects just‑in‑time credentials, and logs the action for audit. Throughout the agent’s life, Astrix monitors credential rotation, usage patterns, and privilege changes, revoking access instantly if anomalies appear. When an agent is retired—perhaps after a model is deprecated or a workflow is discontinued—the orchestrator triggers a decommissioning playbook that revokes tokens, removes associated keys, and updates the inventory. This end‑to‑end management reduces the risk of orphaned NHIs that linger as hidden backdoors.
Agentic Threat Detection and Response
“Agentic threat detection and response: Detect and respond to threats such as compromised credentials and out‑of‑scope agent actions.” Even with strong governance, adversaries constantly seek ways to abuse agent privileges. Astrix’s threat‑detection engine correlates telemetry from identity providers, cloud workload protection platforms, and network traffic to spot suspicious behavior. For example, if a service account that normally reads only from a specific S3 bucket suddenly attempts to list IAM roles or invoke a Lambda function outside its declared scope, the platform raises an alert in real time. Likewise, stolen API keys or exposed secrets trigger immediate credential invalidation and forced re‑authentication. Response actions are automated wherever possible: quarantining the offending agent, initiating a forensic snapshot, or launching a playbook that notifies the security operations center (SOC) and prompts manual review. By blending detection with rapid, policy‑driven remediation, Astrix aims to close the gap between agent capability growth and traditional security models.
Challenges and Industry Context
The urgency behind Cisco’s move is underscored by stark industry figures. As Bailey wrote in a recent blog, “As organizations race to adopt agentic technologies, security teams are under real pressure to enable their teams while securing the use of agents. But agent capabilities are advancing faster than most security models, creating a widening gap in visibility, governance, and response.” Cisco’s AI Readiness Index reveals that only 24% of organizations can control agent actions with proper guardrails and live monitoring, and just 31% feel fully capable of securing their agent AI systems. These statistics illustrate a market where most enterprises are flying blind despite rapid adoption of generative AI, robotic process automation, and multi‑agent orchestration platforms. The sheer scale of the problem is highlighted by Astrix co‑founders Alon Jackson and Idan Gour, who noted, “Agents and other NHIs outnumber humans 100:1, they remain under the radar, and they create the biggest blind spot in our identity perimeter.” This imbalance means that traditional identity‑and‑access management (IAM) tools, built around human users, are insufficient for protecting the proliferation of machine‑driven actors.
Astrix’s Role and the Cisco Acquisition
Astrix positioned itself as the go‑to platform for precisely this dilemma. Its founders proclaimed, “Astrix became the platform security teams turn to when they need to discover, govern, and protect every agentic and non‑human identity across their environment, from provisioning to decommissioning, from policy enforcement to real‑time threat detection. Joining Cisco means Astrix now has the scale, the reach, and the platform to bring agentic and NHI security to organizations worldwide.” By joining Cisco, Astrix gains access to a global sales force, deep integration points with Cisco’s networking, security, and observability stacks, and the financial backing needed to accelerate product development. Cisco, in turn, bolsters its security portfolio with a niche but rapidly growing capability that complements its existing offerings such as Duo Secure Access, SecureX, and Splunk Enterprise Security. The acquisition signals Cisco’s intent to move beyond perimeter‑centric defenses and embed identity‑centric controls directly into the fabric of AI‑driven workloads.
Galileo Acquisition and AI Observability Synergy
Astrix is not Cisco’s only recent foray into AI‑centric security. In April, Cisco announced plans to acquire AI observability firm Galileo Technologies. Galileo’s platform delivers real‑time observability and guardrails for the development of multi‑agent systems, providing visibility into how agents interact, where latency spikes occur, and whether policy violations emerge during model training or inference. Cisco noted that the technology will strengthen its Splunk observability portfolio and bring improved AI agent monitoring capabilities, real‑time visibility, and protection to the agent development lifecycle. Together, Astrix and Galileo give Cisco a two‑pronged approach: Astrix handles the identity‑centric governance and threat response for deployed agents, while Galileo offers deep insight into agent behavior during creation and testing. This combination enables organizations to secure agents from cradle to grave—spanning development, deployment, and runtime.
Future Outlook
The combined strength of Astrix’s discovery‑governance‑response stack and Galileo’s observability layer positions Cisco to address a critical security gap that is likely to widen as autonomous agents become more prevalent in areas such as autonomous supply chains, AI‑driven customer service bots, and self‑optimizing network functions. Analysts predict that by 2027, NHIs could constitute over 90% of all identity entities in large enterprises, making agent‑centric IAM a core component of any zero‑trust architecture. Cisco’s early investments suggest it aims to become a trusted provider of end‑to‑end AI security, offering enterprises the confidence to innovate with agents without sacrificing compliance or resilience. As the market matures, we can expect tighter integrations between Cisco’s networking gear (e.g., Secure Firewall, SD‑WAN) and its agent‑security services, enabling policy enforcement at the point of traffic ingress and egress, further reducing the attack surface.
Conclusion
Cisco’s acquisition of Astrix Security—and its earlier move to bring Galileo into the fold—reflects a strategic recognition that the rise of AI agents introduces a new class of identity risk that traditional tools cannot manage. By providing capabilities to discover, govern, manage lifecycles, and detect threats against non‑human identities, Astrix fills a vital blind spot in the identity perimeter. Combined with Galileo’s real‑time observability for agent development, Cisco now offers a comprehensive suite that spans the entire agent lifecycle. As Bailey warned, the gap between agent capability and security readiness is growing; Cisco’s recent actions suggest it is determined to close that gap, giving organizations the map, guardrails, and response mechanisms they need to secure the agentic future.
https://www.networkworld.com/article/4166695/cisco-grabs-astrix-to-secure-ai-agents.html