The new solution is designed to give organisations visibility into approved and shadow AI use and identify unusual activity by AI agents.

Consider a possible scenario in the workplace: an AI agent summarises internal documents, retrieves information from a business application and uses connected tools to trigger further actions. Such workflows can help employees work more efficiently. They also raise security questions: What data did the agent access? Which user or system permissions did it use? And was any information transferred to an external service?

As organisations adopt AI agents, tracking these activities is becoming more challenging. Agents can operate under user or service accounts and use the permissions available to them. This can put sensitive data, applications and critical systems at risk, particularly when activity spans different models, applications, tools and external services.

Exeon Analytics has introduced Exeon.AgentGuard, a security solution designed to detect risks associated with AI use and agent activity. It is intended to provide visibility into approved and unapproved AI applications, autonomous agents and data flows between models, applications, connected tools and external resources. Security teams can use it to investigate potential data leaks and identify unusual behaviour.

Monitoring AI agent behaviour

According to Exeon, AgentGuard uses existing data sources and does not require additional software agents in the IT environment or complex implementation. The solution is designed to identify deviations from expected AI agent behaviour. Risks it addresses include unauthorised data transfers, misuse of permissions and potentially harmful actions on critical systems.

“AI agents are fundamentally changing how businesses work. Whether operating autonomously or on behalf of employees, they can access sensitive data and perform actions using legitimate permissions – often beyond the visibility of traditional security measures,” says Gregor Erismann, CEO of Exeon Analytics. “Exeon.AgentGuard shows what AI agents are doing and detects unexpected behaviour. Our aim is to help businesses benefit from AI-driven innovation without compromising security.”

Extending threat detection

With AgentGuard, Exeon is extending its behavioural analytics to cover AI use and autonomous agent activity. Exeon.NDR detects network threats through network metadata analysis, while Exeon.UEBA monitors the behaviour of users, identities and business-critical applications. AgentGuard adds monitoring of external AI services and agent activity across models, applications, connected tools and external resources.

The launch reflects an expanding attack surface that now encompasses networks, cloud infrastructure, human and non-human identities, applications and AI agents. Exeon says the additional visibility will help security teams identify and investigate complex threats while retaining control over how and where corporate data is processed.

Exeon Analytics is presenting AgentGuard and other product updates at it-sa Expo&Congress, taking place in Nuremberg from 27 to 29 October 2026.