New Agent Behavior Verification framework helps enterprises assess AI agent readiness before deployment
Exabeam has introduced Agent Behavior Verification (ABV), a new security discipline designed to help organizations evaluate whether AI agents are configured, authorized, and governed appropriately before they are deployed into production environments.
Alongside the launch, Exabeam announced the release of Praxen, an open-source tool that operationalizes the ABV framework, enabling enterprises, developers, and security teams to assess whether AI agents can safely perform their intended roles while remaining within defined security and governance boundaries.
As organizations increasingly deploy autonomous AI agents to access systems, execute workflows, and make decisions, security leaders face growing concerns around agent permissions, governance, and operational risk. While traditional approaches such as vulnerability assessments, penetration testing, and runtime monitoring help identify security weaknesses, they often do not verify whether an AI agent’s behavior aligns with its authorized purpose before deployment.
“As AI agents become digital workers, organizations need confidence that agents have the right permissions, controls, and boundaries before they enter production.” — Steve Wilson, Chief AI Officer, Exabeam
According to Exabeam, Agent Behavior Verification addresses this challenge by evaluating AI agents as complete systems. The framework assesses an agent’s permissions, tools, configurations, memory, integrations, and operating environment against a predefined policy contract known as a “remit,” which outlines what the agent is allowed to do and the boundaries within which it must operate.
Praxen analyzes these factors and identifies gaps between intended and actual behavior, providing actionable recommendations, security findings, and an overall maturity score to help organizations strengthen AI governance and reduce risk.
Steve Wilson, Chief AI Officer at Exabeam and Founder and Co-Chair of the OWASP Gen AI Security Project, said organizations are rapidly moving from AI experimentation to operational deployment. He noted that security teams require greater assurance that AI agents have appropriate controls, permissions, and safeguards before entering production environments.
Released under the Apache 2.0 open-source license, Praxen is designed to encourage industry collaboration and transparency. Exabeam believes the framework will help establish best practices for securing autonomous AI systems and building trust in the growing agentic AI ecosystem.
The launch represents a significant step toward improving AI governance, security, and accountability as enterprises accelerate the adoption of autonomous digital workers.
