Cisco announces intent to acquire Astrix Security to protect the AI agent workforce
The acquisition of Astrix Security by Cisco comes at a time when AI agents are already part of everyday business operations — and that changes a lot from a security standpoint. The announcement, made officially by Cisco, marks another step by the tech giant in a strategy that has been taking shape for months to address the risks that emerge when machines start operating with real autonomy inside corporate environments.
Think about it: today, these agents access data, make decisions, and execute tasks using real credentials, just like any employee would. But unlike a human worker, they operate at machine speed, without breaks, without direct supervision, and often without any formal governance behind them. We are talking about API keys, service accounts, and OAuth tokens — exactly the type of credential that Astrix Security was built to protect.
It is basically a new coworker that nobody officially hired, but who already has access to most of the house. 😅 This scenario had been growing quietly, and Cisco decided to act before things spiraled out of control. The move is part of a broader company strategy to bring the concept of Zero Trust to non-human identities — a territory that, until recently, sat in a blind spot of corporate security tools.
The problem nobody was really watching
For years, corporate security strategies were built with people in mind. Users with logins, passwords, multi-factor authentication, role-based access control — all of that was developed to protect human identities within an IT environment. The problem is that the world changed way too fast, and the tools did not keep up at the same pace. Today, a large portion of the interactions within a digital infrastructure do not come from humans: they come from APIs, automated scripts, software integrations, and more recently, from AI agents that operate autonomously and continuously inside corporate systems.
These agents do not show up on the HR employee roster, but they do show up in access logs. They hold tokens, API keys, and credentials that allow real movement within data environments. And the detail that makes this even more critical is that they rarely go through periodic permission reviews, they have no direct owner, and in many cases, they accumulate privileges that were granted at some point and never revisited. It is a massive attack surface that grew almost organically, without anyone having planned for it.
According to the Cisco AI Readiness Index, only 24% of organizations can control agent actions with proper guardrails and real-time monitoring. On top of that, only 31% feel fully capable of protecting their agentic AI systems. These numbers show that the gap between the technological capability of agents and the readiness of security teams keeps widening — and malicious actors or AI-based threat models like Mythos only make this landscape even more challenging.
Astrix Security was one of the first companies to treat this problem with the seriousness it deserves. Founded five years ago to focus specifically on managing and protecting non-human identities (NHI — Non-Human Identity), the company developed technology capable of mapping, monitoring, and controlling access by non-human entities within complex corporate environments. This includes everything from third-party integrations to the new AI agents being deployed at scale across organizations. When Cisco looked at this portfolio, the logic behind the acquisition became pretty clear.
Receive the best innovation content in your email.
All the news, tips, trends, and resources you're looking for, delivered to your inbox.
By subscribing to the newsletter, you agree to receive communications from Método Viral. We are committed to always protecting and respecting your privacy.
What Astrix Security brings to the table
The Astrix platform is not generic. It was built from scratch to solve specific security problems related to non-human identities, and that is reflected in the capabilities the company offers. Here is a breakdown of what is included in this package:
- Discovery and governance for AI agents: the platform creates a complete map of agentic activity within an organization, verifying access policies, resolving digital hygiene issues, reducing attack surfaces, and preventing compliance violations.
- Agent access and lifecycle management: it enables managing AI agents and their non-human identities from provisioning all the way through full deactivation, covering the entire lifecycle of these entities.
- Agentic threat detection and response: it identifies and responds to threats such as compromised credentials and agent actions that go beyond authorized scope.
- Centralized secrets management: unified management of secrets and sensitive credentials distributed across vaults and cloud environments.
Beyond the technology, Cisco highlighted the importance of the team that comes along with this acquisition. The Astrix Security team is made up of security specialists who have been focused on this space from the beginning, bringing deep technical knowledge and hands-on experience that would be difficult to replicate internally in the short term.
Zero Trust beyond the human user
The concept of Zero Trust is not new — it has existed for over a decade as a security philosophy. The core idea is simple: never trust, always verify. No user, device, or system should have guaranteed access just because it is inside the corporate network perimeter. Everything needs to be authenticated, authorized, and continuously monitored. But in practice, most Zero Trust implementations were designed with human users in mind. Non-human identities were left out of that equation for a long time, and that gap became an increasingly attractive exploitation vector for sophisticated attacks.
With Astrix technology integrated into Cisco’s portfolio, the goal is to expand Zero Trust to also cover this universe of automated entities. In practice, that means having complete visibility into which agents exist within the environment, what permissions they hold, whether those permissions still make sense, and whether the behavior of those agents falls within expected parameters. It is an approach that goes far beyond simply blocking unauthorized access — it creates a governance layer that simply did not exist for non-human identities in most organizations until now.
At the RSA Conference, Cisco had already presented its vision for Zero Trust for the agentic workforce, combining identity discovery and management, access policy enforcement, and runtime behavioral protection to govern how agents operate within corporate systems. The Astrix integration is the missing piece to turn that vision into operational reality.
The timing of this integration could not be more strategic. With the adoption of platforms like Microsoft Copilot, Salesforce agents, Google Workspace automations, and dozens of other AI-based solutions being deployed simultaneously across enterprises, the number of non-human identities operating in corporate environments has grown exponentially over the past two years. Each of these agents represents a potential access point, and without a robust Zero Trust framework covering these entities, the security posture of any organization is compromised, regardless of how sophisticated the other layers of protection may be.
How Cisco has been building its AI security strategy
The acquisition of Astrix did not happen in a vacuum. Cisco had already been making consistent moves to position itself as a leader in security for environments that use artificial intelligence at scale. AI Defense, for example, helps organizations build AI applications securely, protecting models and agents developed in-house. The Zero Trust access architecture was extended with capabilities from Cisco Secure Access and agentic identity features in Duo.
The company also invested in broader foundations for AI security: open-source models, tools for model analysis and protection, guidelines for secure AI development, MCP gateways, and new capabilities across the firewall portfolio to recognize and inspect traffic generated by artificial intelligence agents.
Another relevant project is Glasswing, which has been used to harden Cisco’s own products against increasingly sophisticated threat agents. The company also doubled down on resilient infrastructure and launched Live Protect, an eBPF-based solution that brings protection directly to the network infrastructure, closing critical gaps in defense against vulnerabilities.
And then there is Splunk, of course. Through the platform acquired by Cisco, SOC (Security Operations Center) automation is being accelerated so organizations can respond to incidents at the same speed agents operate. Observability was also expanded to cover AI environments, including the recent acquisition of Galileo, which reinforces that capability.
How it all connects within the Cisco ecosystem
Cisco’s plan is to integrate Astrix Security capabilities into Cisco Identity Intelligence, strengthening visibility and context around identities within the company’s security platform. Additionally, the intent is to extend these capabilities to Zero Trust access solutions, including Cisco Secure Access and Duo Identity and Access Management.
In practice, customers will be able to discover, authenticate, and authorize agentic identities, as well as detect and respond to anomalous behaviors using these same tools. Cisco’s differentiator in this context is the visibility the company has across multiple layers — identity, network, application, and infrastructure. It is not just about knowing what an agent is, but understanding how it behaves over time and across different contexts.
All of this intelligence feeds into Splunk or whatever other SIEM the organization uses, giving security teams a unified view of agent activity with the context they need to investigate and respond at the right speed.
What actually changes for businesses
For those working in corporate security or leading IT teams, the integration of Astrix into Cisco’s portfolio represents a concrete shift in how organizations will be able to manage the risk associated with AI agents. The Astrix platform already offered capabilities like automatic discovery of all integrations and non-human identities connected to the environment, real-time risk analysis, and remediation recommendations. With Cisco’s resources and scale behind it, this technology will reach a much larger customer base and go through advancements that would have been slower for an independent startup.
On top of that, the acquisition positions Cisco in a meaningful way within a market that is consolidating rapidly. Non-human identity management — or NHI — has become a specific segment within the security market, with investors and major companies paying increasing attention. Cisco was not the only one moving in this direction, but the combination of scale, network infrastructure, and the ability to integrate with other security solutions puts the company in a very competitive position to lead this category.
Cisco also reinforced its commitment to open standards, which ensures integration with different environments and allows the company to serve customers wherever they are in terms of technological maturity and IT architecture. This approach is especially important in a landscape where many companies operate with hybrid and multi-cloud environments, where the complexity of managing identities — human or otherwise — is multiplied many times over.
For businesses already using Cisco solutions on a daily basis, the expectation is that Astrix technology will be gradually integrated into existing platforms, especially within the Cisco Security Cloud ecosystem. This could significantly simplify the centralized management of human and non-human identities in a single dashboard, reducing the operational complexity that many security teams face today when trying to cover all these vectors with separate and disconnected tools.
The takeaway from this move is straightforward: corporate security that does not consider non-human identities as part of the Zero Trust model is operating with a serious gap — and that gap will only grow as AI agents become increasingly present and autonomous within organizations. 🔐
