Google Cloud open-sources tool to manage shadow AI

Coverage from SDx Central indicates that Google Cloud has open-sourced a new Kubernetes tool designed to help organizations manage and detect instances of shadow AI within their cloud environments. The tool, named k8s-aibom, aims to provide transparency and auditability for artificial intelligence operations.

K8s-aibom functions as a controller that surfaces a bill of materials (BOM) for AI operations, exposing them across a user's cloud estate. It monitors AI inference, training, agents, evaluations, and retrieval-augmented generation (RAG) frameworks. By utilizing Cluster APIs and container platforms, the tool automatically scans running AI environments and generates CycloneDX Machine Learning Bills of Materials (AI/ML-BOMs), a standard focused on AI transparency. Google states that k8s-aibom aids auditors in tackling shadow AI by minimizing security measure bloat and developer friction. It establishes an audit-grade evidence trail without modifying existing configurations or injecting code into the Linux kernel.

The tool operates in a read-only manner, fitting into GitOps workflows for change detection and alerts. It is compliant with the EU AI Act, NIST AI Risk Management Framework, and ISO/IEC 42001, using a structured taxonomy to differentiate engineering intent from machine inference, thereby building trust during audits.

Source: SDx Central