Tenable Launches Exchange Inspector to Vet Open-Source AI Security Agents
Tenable introduces a new vetting process using OpenAI models and human review to validate the security of community-built AI agents before deployment.
Key points
- Tenable uses automated checks, frontier model assessments, and human verification to vet AI agents.
- The process screens for 15 security issue types, including prompt injection and excessive permissions.
- Three agents, including two from Tenable and one from Splunk, have passed the initial review.
Tenable has launched the Exchange Inspector, a new vetting mechanism designed to evaluate the security posture of open-source AI agents before they are deployed in security operations. According to Tenable, this initiative addresses the rapid integration of community-built AI tools into SOC workflows by establishing a standardized review process for listings on the CyberAgents Exchange.
The Exchange Inspector subjects each candidate tool to a three-stage evaluation pipeline to earn a vetted tag. This pipeline includes an automated check, an assessment using frontier AI models, and final human verification by Tenable security researchers. According to Tenable, this multi-layered approach ensures that agents meet rigorous security standards before being recommended for operational use in enterprise environments.
Vetting Methodology and Scope
Tenable employs Tenable One AI Exposure to screen for specific threats such as prompt injection vulnerabilities and exposed secrets within the agent code. The company also utilizes OpenAI GPT Cyber models to assess the overall code structure and threat model of each submission. According to Tenable, these AI-driven assessments provide a foundational layer of security analysis that identifies common coding flaws and potential attack vectors early in the review process.
Security researchers then verify the runtime behavior of each agent in a clean, isolated environment to confirm that the automated findings are accurate and that the tool behaves as expected. This human-in-the-loop verification step is critical for detecting subtle issues that automated scans might miss. According to Tenable, this comprehensive review tests for 15 distinct types of security issues across three layers of the software stack.
The security issues identified during vetting range from conventional vulnerabilities like server-side request forgery (SSRF) and path traversal to agent-specific risks such as excessive permissions and memory poisoning. According to Tenable, addressing these diverse threat categories is essential because AI agents often require broad access to data and systems, making them high-value targets for attackers if not properly secured.
Current Listings and Operational Impact
The CyberAgents Exchange currently features three listings that have successfully passed the Exchange Inspector vetting process. Two of these vetted agents were built by Tenable, while the third was developed by Splunk. According to Tenable, these initial successes demonstrate that community contributions can meet high security standards when subjected to rigorous evaluation and proper engineering practices.
Contributors to these vetted tools report significant operational efficiencies, including accelerated threat hunting, faster remediation triage, and improved cloud posture investigation capabilities. Specifically, the SOC Hunter agent has reduced hunt times by 75 percent for Tenable’s internal security team. According to Tenable, these efficiency gains highlight the potential value of AI agents in SOC workflows when they are trusted and secure.
Security teams are increasingly deploying AI agents to triage alerts, correlate threat intelligence, and prioritize vulnerabilities across their environments. However, the speed of adoption has outpaced the development of standardized security reviews for these tools. According to Tenable, it is critical for organizations to understand exactly what these AI agents will do before deploying them to avoid introducing new risks into their security infrastructure.
What should you do?
Organizations deploying AI agents should verify that the tools undergo independent security assessments before integration. Security teams should monitor for excessive permissions requested by AI agents and restrict access to only necessary data sources. Watch for signs of prompt injection or unexpected behavior in agent logs. Ensure that any AI tool used in SOC workflows has been vetted for known vulnerabilities like SSRF.
Background: Prompt injection attacks
Look for sudden shifts in tone, refusal to follow system constraints, or output that repeats hidden instructions. These signs indicate the model is processing injected commands rather than your intended query.
Read the full guide: Prompt Injection Warning Signs: How to Spot AI Manipulation
What to do and how to stay safe: Tenable
- Verify that AI agents undergo independent security assessments before integration into SOC workflows.
- Monitor for excessive permissions requested by AI agents and restrict access to necessary data sources.
- Watch for signs of prompt injection or unexpected behavior in agent logs during initial deployment.
- Ensure that any AI tool used has been vetted for known vulnerabilities like server-side request forgery.
General security guidance from the Threat Times newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
How many AI agents have passed the Exchange Inspector vetting so far?
According to Tenable, three agents have passed the vetting process, including two built by Tenable and one by Splunk.
What types of security issues does the Exchange Inspector check for?
The review tests for 15 types of issues, including prompt injection, exposed secrets, SSRF, path traversal, and memory poisoning.
What tools does Tenable use to assess the AI agents?
Tenable uses Tenable One AI Exposure and OpenAI GPT Cyber models for automated and model-based assessments.
