TrustX ARC Framework Classifies Agentic AI System Risks

Hannah M. Liu, Rhea Saxena, Shiv Asthana· July 13, 2026 View original

▶ The 2-minute explainer

Summary

The TrustX Agent Risk Classification (ARC) Framework provides a structured method for risk-tiering internally created agentic AI systems, addressing the gap in general-purpose AI risk frameworks. It uses a twelve-dimension scoring rubric, a GPA + IAT classification model, and a five-level autonomy framework to produce a three-tier governance output with control recommendations.

This paper introduces the TrustX Agent Risk Classification (ARC) Framework, a new instrument designed to help organizations classify and govern the risks associated with agentic AI systems. As these autonomous AI agents become more prevalent across enterprises and public sectors, existing general-purpose AI risk frameworks often fall short in addressing their unique complexities. ARC aims to fill this gap by providing a structured and repeatable methodology applicable to seven distinct types of agentic AI. At its core, the framework features a twelve-dimension scoring rubric that quantifies risk robustly. This rubric is integrated with other established components, including a GPA + IAT classification model and a five-level autonomy framework. The combination of these inputs generates a three-tier governance output, complete with mapped control recommendations. The framework also includes a specialized extension for Coding Assistant agents, acknowledging their specific nuances. ARC is intended for a broad audience, including AI governance practitioners, risk officers, developers, and regulators, and is designed for continuous iteration and improvement.

Why it matters

For organizations deploying or developing agentic AI, this framework provides a critical tool for systematically assessing and managing risks, ensuring responsible AI adoption and compliance with emerging governance standards.

How to implement this in your domain

  1. 1Download and review the interactive TrustX ARC Framework to understand its components and methodology.
  2. 2Identify all agentic AI systems currently in use or under development within your organization.
  3. 3Apply the twelve-dimension scoring rubric to each identified agentic AI system to quantify its risk profile.
  4. 4Utilize the framework's classification models to assign a risk tier and implement the corresponding control recommendations.
  5. 5Establish a regular review process to re-evaluate agentic AI systems using ARC as they evolve or new ones are introduced.

Who benefits

BFSIGovernmentTechnologyHealthcareLegal

Key takeaways

  • The TrustX ARC Framework helps classify and govern risks of agentic AI systems.
  • It uses a twelve-dimension rubric and integrates existing AI governance models.
  • The framework produces a three-tier governance output with control recommendations.
  • It is designed for AI governance practitioners, risk officers, developers, and regulators.

Original post by Hannah M. Liu, Rhea Saxena, Shiv Asthana

"arXiv:2607.09586v1 Announce Type: new Abstract: The proliferation of agentic AI systems across enterprise and public-sector contexts has outpaced the capacity of general-purpose AI risk frameworks to classify and govern them. In this paper, we introduce the TrustX Agent Risk Clas…"

View on X

Originally posted by Hannah M. Liu, Rhea Saxena, Shiv Asthana on X · view source

Want to go deeper?

Turn these trends into skills with Learnijoy's hands-on AI & tech courses.

Explore courses