Security intelligence and management solutions company Exabeam Inc. today announced an expansion of its Agent Behavior Analytics platform that adds detection and monitoring of artificial intelligence agent activity across tools such as OpenAI Group PBC’s ChatGPT, Microsoft Corp.’s Copilot and Google LLC’s Gemini.

The expansion seeks to address the issue whereby, without direct visibility into how employees are using AI assistants, organizations cannot baseline normal AI behavior, investigate possible misuse, or detect emerging agentic insider threats.

The new support to detect agent behavior is designed to transform agentic services into rich sources of behavior telemetry that feed directly into Exabeam threat detection, investigation and response workflows.

“AI agents are evolving from simple chatbots into autonomous digital workers,” said Steve Wilson, Exabeam’s chief AI and product officer. “They authenticate, access systems and execute real business processes. When compromised, their activity will often look legitimate.”

But, he added, “guardrails designed to catch prompt injection or hallucinations do not address that risk. Securing digital workers requires deep visibility into baseline behavior and the ability to detect subtle deviations before they become material incidents.”

To address the fast-evolving agentic threats, Exabeam today launched five new capabilities that it says work together to provide complete coverage of the agentic attack surface.

AI behavior baselining builds dynamic behavior profiles for users and their AI agents and tracks patterns across request volumes, token usage, tool invocations, web sessions and outbound activity. When behavior deviates from established norms, Exabeam flags the anomaly to help security teams detect misuse before it escalates.

Prompt and model abuse detection, as the name suggests, detects prompt injection, model manipulation and tool exploitation before attacks escalate. The detection capability is complemented by a new detection library, which is five times larger than the previous version to cover the full threat spectrum, including prompt manipulation and shadow AI activity.

Identity and privilege monitoring governs what agents are allowed to do, and agent lifecycle monitoring provides full visibility into the creation, modification, and usage of AI agents and coverage.

The final capability, coverage for OWASP Top 10 for Agentic AI, sees agent behavior monitored against the Open Worldwide Application Security Project’s agentic AI framework to bring measurable coverage to a threat category that previously lacked a defined framework.

The new capabilities are also accompanied by a set of enhancements across the Exabeam New-Scale and LogRhythm Platforms. The enhancements are designed to improve the day-to-day experience for administrators and security analysts while continuing to deliver visibility and automated response to help teams streamline workflows, reduce alert fatigue and accelerate threat detection.

Photo: Exabeam