EVERYTHING
According to Siliconangle, Exabeam Inc. has launched a major update to its security operations platform, specifically targeting the growing risks associated with autonomous AI agents. The company now provides detection coverage for 90 different AI models, notably adding support for Anthropic PBC’s Claude. This expansion follows a year of continuous development in agent-specific security features aimed at closing visibility gaps in modern enterprise environments.
Detecting machine-speed anomalies
As organizations transition from simple AI experimentation to deploying autonomous agents, security teams face the challenge of monitoring entities that act on behalf of users. Because these agents can invoke tools and access systems rapidly using legitimate credentials, their malicious activity often blends in with normal operations. Exabeam’s new detections are designed to flag several specific types of suspicious behavior:
- Anomalous interactions between human users and AI agents.
- Unauthorized autonomous activity and suspicious prompt behaviors.
- Unusual sequences of tool invocations or abnormal consumption patterns.
- "Denial-of-wallet" indicators, which track excessive resource costs used in attacks.
- Shadow AI usage and unauthorized configuration changes across the network.
Expanded tools and open-source contributions
While Claude is a primary addition to the platform, Exabeam already maintains coverage for major models including OpenAI’s ChatGPT, Google’s Gemini, and Microsoft’s Copilot. The company also integrated the OWASP Top 10 for Agentic AI into its Outcomes Navigator tool, allowing security teams to benchmark their detection capabilities against industry standards.
To further assist detection engineers, Exabeam introduced Nova Rules Creator, which allows users to build correlation rules using natural language. Additionally, the company released Observra, an open-source telemetry library designed to capture and normalize activity across major AI frameworks. This library enriches data with cost and risk signals before routing it to security platforms. "Security teams need visibility not only into human activity, but into how agents behave, interact and make decisions," — Pete Harteveld, Chief Executive of Exabeam.
The update also includes broader infrastructure improvements, such as expanded LogRhythm SIEM integrations across cloud and identity technologies. By providing a comprehensive telemetry layer for both pre-deployment verification and live monitoring, Exabeam aims to standardize how enterprises govern the lifecycle of AI agents.