Linx Security launches tools to govern AI agent actions
The company is expanding its identity governance to track and control what autonomous AI agents do across corporate networks.
The short version
- Linx Security now detects shadow AI agents running on endpoints, in CI/CD pipelines, and within SaaS or cloud environments.
- The new Agent Runtime Control uses the AuthZEN open standard to allow or block tool calls based on risk and identity context.
- Users can create 'Autopilots' by giving agents plain-English missions with specific scopes and human-approval requirements.
- The 'Prompt-to-Workflow' feature converts natural language descriptions into tested, agentic workflows for identity teams.
- Linx customers already manage over 25,000 AI agent identities, which grew 450x in just six months.
Why it matters
As companies adopt autonomous agents, traditional security models designed for humans struggle to keep up with machine-speed actions. These tools help IT teams prevent 'shadow AI' from accessing sensitive data while allowing for safer automation.
The challenge of machine-speed identity
Security teams are facing a new hurdle as AI agents begin to act autonomously across company tools and data. Unlike humans, these agents can be created, renamed, and modified at machine speed, often leaving very little trace for traditional security systems. Because they can react without a specific script, it is often difficult for administrators to predict what an agent will do next.
Many of these agents are considered 'shadow AI,' meaning they run on personal accounts or outside of sanctioned corporate platforms. They can appear in pipelines or AI applications without a clear owner or any formal security review. This lack of visibility creates a significant risk for enterprises trying to maintain control over their digital environments.
Gartner predicts that over 50% of enterprise AI initiatives will stop by 2028 because of these unresolved identity challenges. To address this, Linx Security is moving beyond simple monitoring to a model that treats AI agents as identities that require the same governance as humans. The goal is to understand the context of every action an agent takes rather than just seeing that an action occurred.
Mapping the identity graph
Linx Security uses an Identity Graph to organize the complex relationships between humans, non-human identities (NHIs), and AI agents. This graph places every agent, whether it is officially sanctioned or not, into a single view. It tracks who owns each agent and what specific permissions or access rights that agent carries across the organization.
By putting all identities on one map, the system can weigh the risk of every specific tool call or access change. This context-aware approach allows security teams to see the full picture of an agent's reach. Instead of treating agents as a separate problem, the company integrates them into its existing architecture for managing all types of identities.
The system works in three distinct steps: graphing the identities, deciding on policies based on that graph, and controlling the actions. This ensures that security decisions are made based on the specific risk of the agent's behavior. It provides a way to enforce these rules across whatever gateway the agents are currently using to interact with company data.
Detecting shadow AI and runtime control
One of the new capabilities is Shadow AI Agent Detection, which looks for agents running outside of sanctioned platforms. This includes finding agents hidden in endpoints, cloud environments, and CI/CD pipelines. When a shadow agent is found, it is added to the Identity Graph with information on its owner, potential exploits, and instructions on how to fix the security gap.
To manage how agents behave in real-time, Linx introduced Agent Runtime Control. This feature integrates with the AI gateways that organizations already use, acting as a decision point through the AuthZEN open standard. If a company does not have a gateway, they can use the Linx MCP Gateway to manage these interactions.
Every time an agent tries to make a tool call, the system decides whether to allow or block it. This decision is based on who is behind the agent, the level of risk involved, and what the agent is trying to do. This allows for granular control without requiring companies to replace their existing infrastructure or change their current workflows.
Automating identity workflows
Linx is also introducing tools to help identity teams use agents to do their own work. The 'Autopilots' feature allows teams to give an agent a specific mission described in plain language. Teams can set the scope of the mission and choose which actions the agent can take alone and which ones require a human to provide approval first.
These Autopilots are built using the organization's existing knowledge of identities and access. Before an Autopilot goes live, teams can review the plan and perform a dry-run to see how it will behave. This allows for the creation of governed agents that can handle repetitive tasks while staying within the boundaries set by the security team.
Additionally, the 'Prompt-to-Workflow' tool lets teams describe a process, such as deactivating an agent when an owner leaves the company. Linx then builds the steps for that workflow and shows the underlying code to the user. The system performs a dry-run before the workflow goes live to ensure it functions as expected. As Israel Duanis, CEO of Linx Security, noted, "Enterprises need to know... what every identity is connected to, what it’s doing and why."
Scaling for enterprise growth
The demand for these tools is already showing significant growth as companies rush to adopt AI. Linx Security reports that its customers are already governing more than 25,000 AI agent identities in production environments. This number grew by 450x in just a six-month period, highlighting the rapid shift toward agentic workflows.
By providing a way to manage both sanctioned and unmanaged agents, Linx aims to bridge the gap between human-speed governance and machine-speed actions. The platform ensures that as more agents are deployed, they do not become a source of unmonitored risk. It provides a structured way to scale AI while maintaining a clear audit trail of every action taken.
Moving forward, the focus remains on providing the context necessary for secure AI adoption. By integrating agents into the existing identity framework, companies can move faster with AI without sacrificing security. The goal is to ensure that every autonomous action is tied to a known identity with a clear purpose and set of permissions.
What to watch
- How many other identity vendors will adopt the AuthZEN standard for agent governance.
- The percentage of enterprise AI projects that actually successfully clear identity hurdles by 2028.
Try it
Check if your organization has a formal policy for 'shadow AI' or if employees are using personal AI accounts for work tasks.
Sources and links
First reported by Help Net Security. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.
Liked it? Say why
Add a line for other readers.
No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.