Machine-Readable No-Learn Protocols and Fiduciary Governance Under the Extraterritorial Brussels Effect: Technical Specifications, Defensive Publication, and Board Liability Interlocks for Sovereign Intellectual Property

Abstract: Frontier generative artificial intelligence (AI) developers increasingly engage in regulatory arbitrage, relocating corporate entities to permissive jurisdictions while ingesting public intellectual output via automated web crawling, lossy parameter extraction, and model distillation. However, under Article 2(1)(c) of the EU Artificial Intelligence Act (Regulation (EU) 2024/1689), the "Brussels Effect" exerts unavoidable extraterritorial jurisdiction whenever model outputs are utilized within the European single market. Concurrently, Article 53(1)(c) mandates General-Purpose AI (GPAI) providers to respect machine-readable opt-outs under Article 4(3) of the CDSM Directive (Directive (EU) 2019/790). This treatise formalizes a dual-layer technical and statutory defense architecture for independent research sovereignty: (1) The Machine-Readable No-Learn Protocol (NLL-v1.0-2026), integrating cryptographic hashing, persistent digital object identifiers (DOI) anchored on CERN Zenodo, and standardized XMP/HTTP metadata headers that cryptographically signal automated opt-outs, invalidating fair use defenses; (2) Defensive Publication as Prior Art, leveraging immutable Zenodo timestamps to deterministically destroy patent novelty (35 U.S.C. § 102 / EPC Art. 54) against subsequent corporate enclosure; and (3) Fiduciary Governance Interlocks, linking unauthorized model training directly to corporate director liability under the Delaware Caremark Standard. We establish that willful violation of machine-readable opt-outs constitutes bad-faith governance failure, providing sovereign creators with an actionable, verifiable enforcement shield against predatory algorithmic extraction.