Summary
This paper introduces the TrustX Agent Risk Classification (ARC) framework, designed to help organizations govern internal agentic AI systems whose adoption has outpaced the capabilities of general-purpose AI risk frameworks. ARC applies to seven types of agentic AI systems and builds on existing AI governance frameworks. At its core is a twelve-dimension scoring rubric that quantifies risk. Combined with the GPA+IAT classification model and a five-level autonomy framework, ARC produces three-tier governance outputs along with mapped control recommendations. The paper also includes a dedicated extension for coding assistants, addressing the nuances of this specific class of agentic systems. ARC is aimed at AI governance practitioners, risk officers, developers, and regulators, and is intended to be iteratively refined. Authored by Hannah M. Liu, Rhea Saxena, and Shiv Asthana, the paper is available on arXiv as 2607.09586.
Overview
Research area: AI / Governance
Authors: Hannah M. Liu, Rhea Saxena, Shiv Asthana
Published: 2026-07-10
arXiv: 2607.09586
Abstract (translated)
The adoption of agentic AI systems within enterprises has outgrown the classification and governance capabilities of general-purpose AI risk frameworks. This paper introduces the TrustX Agent Risk Classification (ARC) framework, which can be applied to seven types of agentic AI systems and is grounded in existing AI governance frameworks.
At the core of ARC is a twelve-dimension scoring rubric that quantifies risk. Combined with the GPA+IAT classification model and a five-level autonomy framework, the approach produces three-tier governance outputs along with mapped control recommendations.
The paper also includes a dedicated coding-assistant extension to address the nuances of this particular category of agentic systems.
ARC is intended for AI governance practitioners, risk officers, developers, and regulators, and will continue to be iteratively refined.
---
*Auto-collected on 2026-07-14.*
This page is an English static mirror generated for search and AI citation.
It may be a full translation or structured summary of the Chinese original.
Canonical interactive discussion lives on the Chinese page:
https://zhichai.net/topic/178395126