Exabeam, a cybersecurity company that builds behavioral analytics and threat detection software for enterprise security operations, has expanded its security platform with new capabilities designed to give security teams visibility into how autonomous AI agents behave inside their environments. The update doubles Exabeam's AI-focused behavioral detections from 45 to 90 and adds Anthropic Claude to its list of supported AI platforms, which already included OpenAI ChatGPT, Google Gemini, Microsoft Copilot and GitHub Copilot.
The expansion addresses a detection gap that has grown as AI agents have moved from isolated experiments into production environments. Unlike conventional applications, AI agents access systems and invoke tools using legitimate credentials, which means their activity is difficult to distinguish from authorized human or automated behavior through standard security monitoring. Exabeam's approach, which the company calls Behavior Intelligence, applies behavioral analytics to both human and agent activity to surface deviations from established patterns.
The 90 detections now cover suspicious prompt activity, unauthorized autonomous agents, abnormal AI usage patterns, shadow AI deployments, configuration changes and what Exabeam calls Denial of Wallet attacks, a category of abuse in which AI services are exploited to generate excessive compute costs. The company has also aligned its detection coverage with the OWASP Top 10 for Agentic AI, a framework that catalogs security risks specific to AI agent deployments, giving security teams a way to map their existing controls against that list and identify gaps.
Pete Harteveld, CEO of Exabeam, described the shift driving demand for the new capabilities: "Organizations are rapidly moving from AI experimentation to autonomous AI agents operating across the enterprise. Security teams need visibility not only into human activity, but into how agents behave, interact and make decisions."
Alongside the platform updates, Exabeam has released two open-source tools. Observra is a telemetry library that captures and standardizes activity data from AI agents across multiple frameworks and feeds it into security operations platforms. Praxen, previously announced, handles Agent Behavior Verification, helping organizations confirm that AI agents are configured and governed correctly before they are deployed into production. Together, the two libraries are positioned as a pre-deployment and runtime complement to Exabeam's detection capabilities.
Other updates include AI-powered rule creation using natural language input, automated case correlation for faster threat investigation, phishing email ingestion, expanded cloud data collection and updated reporting dashboards.
