A Typed Claim-Governance Ontology for Neuroscience and Self-Aware Networks Research

This revised preprint presents an executable governance layer for a heterogeneous neuroscience and Self-Aware Networks research graph. It separates source identity, evidence class, claim status, editorial verdict, privacy, and allowed actions so that retrieval does not silently become endorsement. The public package includes a synthetic twenty-four-case benchmark, twelve source identities, fifteen relation classes, ten exact-set competency questions, a machine-readable schema, deterministic policy code, mutation tests, and reproducible results. A privacy-only baseline admits ten false promotions, while the governed rule matches the author-coded fixture. That result demonstrates contract behavior on constructed cases only. It does not establish ontology completeness, independent annotation reliability, neuroscience truth, or empirical validation of SAN, PWD, or NAPOT.