AAGATE: A NIST AI RMF-Aligned Governance Platform for Agentic AI

Ken Huang, Kyriakos “Rock” Lambros, Jerry Huang, Yasir Mehmood, Hammad Atta, Joshua Beck, Vineeth Sai Narajala, Muhammad Zeeshan Baig, Muhammad Aziz Ul Haq, Nadeem Shahzad, Bhavya Gupta · 2025

Read AAGATE: A NIST AI RMF-Aligned Governance Platform for Agentic AI by Ken Huang, Kyriakos “Rock” Lambros, Jerry Huang, Yasir Mehmood, Hammad Atta, Joshua Beck, Vineeth Sai Narajala, Muhammad Zeeshan Baig, Muhammad Aziz Ul Haq, Nadeem Shahzad, Bhavya Gupta (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2510-25863.txt.

Introduction

Ken Huang leads ten co-authors in presenting a governance platform for autonomous AI agents used in real workplaces. Agents that act at machine speed and improvise can break the assumptions behind standard software security tools. Their platform, AAGATE, is built to run on Kubernetes, a system for managing container software, and it follows the national AI risk framework from NIST.

Huang and the team attach a specialized method to each risk step, from threat modeling to red team testing. The platform adds a zero trust service mesh, a policy engine that can explain its choices, behavior tracking, and accountability hooks. Extra modules cover identity rights, logic layer attacks, and signs of cognitive decline in agents.

Catalog record

CREATORKen Huang, Kyriakos “Rock” Lambros, Jerry Huang, Yasir Mehmood, Hammad Atta, Joshua Beck, Vineeth Sai Narajala, Muhammad Zeeshan Baig, Muhammad Aziz Ul Haq, Nadeem Shahzad, Bhavya Gupta
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE40 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2510-25863

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

AAGATE: A NIST AI RMF-Aligned Governance Platform for Agentic AI Ken Huang, Kyriakos “Rock” Lambros, Jerry Huang, Yasir Mehmood, Hammad Atta, Joshua Beck, Vineeth Sai Narajala, Muhammad Zeeshan Baig, Muhammad Aziz Ul Haq, Nadeem Shahzad, Bhavya Gupta Abstract This paper introduces the Agentic AI Governance Assurance & Trust Engine (AAGATE), a Kubernetes-native control plane designed to address the unique security and governance challenges posed by autonomous, language-model-driven agents in production. Recognizing the limitations of traditional Application Security (AppSec) tooling for improvisational, machine-speed systems, AAGATE operationalizes the NIST AI Risk Management Framework (AI RMF). It integrates specialized security frameworks for each RMF function: the Agentic AI Threat Modeling MAESTRO framework for Map, a hybrid of OWASP’s AIVSS and SEI’s SSVC for Measure, and the Cloud Security Alliance’s Agentic AI Red Teaming Guide for Manage. By incorporating a zero-trust service mesh, an explainable policy engine, behavioral analytics, and decentralized accountability hooks, AAGATE provides a continuous, verifiable governance solution for agentic AI, enabling safe, accountable, and scalable deployment. The framework is further extended with DIRF for digital identity rights, LPCI defenses for logic-layer injection, and QSAF monitors for cognitive degradation, ensuring governance spans systemic, adversarial, and ethical risks. I Introduction Agentic AI systems don’t sit quietly in a corner—they browse, write code, spin up sub-agents, hit production APIs, and do it all at machine speed. This power is intoxicating for engineers and terrifying for security and compliance teams. A single stray prompt or hallucinated shell command can leak customer data, rack up cloud bills, or rewrite infrastructure. Traditional AppSec tooling, designed for predictable, deterministic applications, was never built for reasoning machines that improvise. These systems also face new vulnerabilities that traditional frameworks can’t address. For instance, Logic-layer Prompt Control Injection (LPCI) is a covert attack hidden in tools and memory atta2025logic . Cognitive degradation is another issue, causing agents to exhibit unstable behavior (QSAF) qsaf2025 . Additionally, the misuse of digital identity likeness necessitates rights-based governance (DIRF) to manage the risk dirf2025 . This new reality demands a new paradigm: continuous, automated, and explainable governance that is as dynamic as the agents it oversees. The NIST AI Risk Management Framework (AI RMF) nist2023 provides a voluntary, structured, and flexible foundation for addressing the multifaceted risks of AI systems. It organizes the practice of AI risk management into four key functions: • Govern: Cultivating a culture of risk management, establishing clear roles, responsibilities, and accountability structures. • Map: Establishing the context in which an AI system operates and identifying the full spectrum of potential risks within that context. • Measure: Employing quantitative and qualitative tools to analyze, assess, track, and monitor identified risks. • Manage: Allocating resources to treat prioritized risks, including mitigation, transfer, avoidance, or acceptance, and implementing response and recovery plans. While the NIST AI RMF provides the essential “what” and “why” of AI governance, it remains non-prescriptive about specific technical implementations. There is a critical need for a practical platform that translates these high-level functions into concrete architectural patterns and engineering controls. Such a platform must not only instantiate RMF functions with existing frameworks (MAESTRO, AIVSS, SSVC, CSA Red Teaming), but also account for emerging research directions in systemic stability (QSAF), covert injection (LPCI), and identity governance (DIRF). This paper introduces the Agentic AI Governance Assurance & Trust Engine (AAGATE), a novel platform designed to be that implementation. AAGATE operationalizes the NIST AI RMF by integrating a minimum set of specialized frameworks into its core architecture to address each function directly: • To Map Risk , AAGATE’s architecture embodies the principles of the Cloud Security Alliance’s Agentic AI Threat modeling MAESTRO framework maestro2025 , providing a layered, multi-agent view and funneling all side-effects through a single chokepoint for comprehensive visibility and threat modeling. • To Measure Risk , AAGATE utilizes a hybrid approach. It generates security signals that can be quantified using the OWASP AI Vulnerability Scoring System (AIVSS) aivss2025 and uses a decision-making logic inspired by the SEI Stakeholder-Specific Vulnerability Categorization (SSVC) ssvc2019 to prioritize responses. • To Manage Risk , AAGATE’s continuous monitoring, incident response, and containment capabilities are designed to implement the proactive, adversarial mindset championed by the Cloud Security Alliance’s (CSA) Agentic AI Red Teaming Guide csa2025 . In presenting this comprehensive platform and its underlying methodology, this paper makes the following novel contributions to the fields of AI governance and security: • An Architectural Blueprint for Operationalizing the NIST AI RMF: The NIST AI RMF provides a critical, high-level conceptual framework for AI risk management, but a significant gap exists between its principles and practical, day-to-day engineering. AAGATE is presented as the first open-source, Kubernetes-native reference architecture that provides a concrete, end-to-end implementation of the RMF’s Govern, Map, Measure, and Manage functions. It serves as a testable blueprint that organizations can adopt and adapt, moving the industry from abstract discussions to tangible implementation. • A Prescriptive, Integrated Toolchain for the AI RMF Functions: While the NIST RMF is intentionally tool-agnostic, this flexibility can be a barrier to adoption for organizations seeking a clear starting point. This paper proposes a novel, minimal, and cohesive set of best-in-class, specialized frameworks to fulfill each RMF function. By explicitly assigning CSA’s MAESTRO to Map, a hybrid of OWASP AIVSS and SEI SSVC to Measure, and the CSA Agentic AI Red Teaming Guide to Manage, we present a strong, defensible hypothesis for what constitutes a complete and practical agentic AI governance stack. This integration synthesizes disparate efforts from leading security organizations into a single, interoperable system. • Novel Architectural Patterns for Agentic AI Governance and Containment: Beyond integrating existing frameworks, AAGATE introduces several novel architectural patterns specifically designed to manage the risks of autonomous agents: – Continuous, Automated Internal Red Teaming: The Janus Shadow-Monitor-Agent (SMA) introduces a new paradigm for runtime assurance. Instead of periodic, manual red teaming, the SMA provides a pre-execution, real-time “shadow” evaluation of an agent’s planned actions, creating a continuous internal challenge to its reasoning and policy adherence. – Mathematically Verifiable Compliance: The use of a ZK-Prover to generate and post on-chain compliance proofs moves beyond traditional, tamper-prone log auditing. It offers a privacy-preserving, mathematically verifiable method for stakeholders to ensure the system is operating within its defined safety and ethical budget. – Purpose-Bound Service Identity: The OAuth Relay mechanism translates the abstract capabilities of an agent into ephemeral, narrowly-scoped, purpose-bound credentials for each specific side-effect. This provides a robust model for least-privilege access in a dynamic, autonomous environment where traditional user-centric consent is impossible. • A Unified Model Bridging Policy, Security, and AI Development: This work bridges the gap between disparate domains that have historically operated in silos. It connects the high-level policy and risk concerns of the NIST AI RMF with the technical threat modeling of MAESTRO, the quantitative risk analysis of AIVSS, the decision-centric prioritization of SSVC, and the adversarial mindset of red teaming. AAGATE provides a common platform and language where compliance officers, security engineers, and AI/ML developers can collaborate, ensuring that governance is not an afterthought but an integral part of the system’s design and operation. • Extension with emerging governance frameworks: AAGATE incorporates DIRF for digital identity rights, LPCI defenses for logic-layer injection, and QSAF monitors for cognitive degradation, extending RMF

Read the complete text