This paper introduces Semantic Hashing, a reframing of hashing as a governed semantic operation rather than a purely cryptographic or structural integrity check. Instead of identifying byte sequences or concrete artifacts, semantic hashes identify equivalence classes of meaning defined by explicit, declared semantic lenses. The core move is intentionally simple and lossy: decide which distinctions are allowed to matter, canonicalize that information, hash it, and allow all other attached meanings to evaporate. When this semantic collapse is governed - explicit, versioned, and inspectable - hashes become instruments of admissibility, drift detection, and temporal anchoring rather than mere checksums. Semantic hashing enables systems to mechanically determine whether something is meaningfully the same without re-interpretation, narrative justification, or re-reasoning. It provides stable semantic addressability, supports refusal without debate, and yields immutable semantic receipts that preserve accountability without freezing evolution. The paper situates semantic hashing as a foundational primitive within governed systems architectures, including its relationship to Governed Semantic Compilation (GSC) and MCIR-style substrates. It emphasizes governance over inference, permission over truth, and declared meaning over emergent interpretation, with applications spanning policy enforcement, program analysis, audit systems, and AI governance.