arXiv cs.AIMonday · May 25, 2026FREE

Ontological Knowledge Blocks: Executable Compliance and Profile-Based Validation for Trustworthy AI Systems

Published on May 25, 2026, a paper on arXiv cs.AI introduces Ontological Knowledge Blocks (OKBs) as a programmable governance infrastructure designed to address the challenges of AI system compliance. Current compliance methods, which are documentation-centric and rely on manual review, are inefficient and do not scale for automated AI services deployed in critical digital infrastructure. OKBs aim to compile regulatory obligations, spanning transparency, accountability, fairness, and traceability, into machine-checkable constraints that operate over structured evidence graphs. This approach seeks to automate the verification process, moving beyond static checklists and manual audits. The paper formally defines an OKB as a 5-tuple, which binds normative obligations to an RDF/OWL concept schema, executable SHACL validation rules, explicit evidence requirements, and PROV-O provenance links. A deterministic regulatory compiler is central to the system, translating structured Intermediate Representation (IR) records into composable Knowledge Block modules. This architecture facilitates profile-based governance reconfiguration, allowing AI systems to adapt to varying compliance profiles without requiring modifications to the underlying service code. Researchers implemented two prototypes and evaluated their effectiveness in an AI-assisted High-Performance Computing (HPC) resource allocation scenario. The evaluation involved 24 validation runs across four distinct governance profiles, demonstrating the system's capability to provide automated and adaptable compliance verification for trustworthy AI systems.