This work introduces Local AI, a systems framework for artificial intelligence that is structurally embedded in place, culture, institutions, and everyday constraints. Rather than treating context as an external prompt or post-processing layer, the framework formalizes locality, governance, and feasibility as first-class system components that shape reasoning, retrieval, and output generation. The paper defines key principles underlying Local AI, including embedded intelligence, constraint-as-signal, and sovereign context, and proposes a reference architecture based on local-first retrieval, structured context frames, privacy-by-design guardrails, and accountability mechanisms. It distinguishes Local AI from related paradigms such as retrieval-augmented generation, edge AI, and federated learning by emphasizing enforceable locality and governance rather than optimization or deployment topology alone. In addition to formal definitions and architectural models, the work discusses policy and ethical implications, positioning Local AI as a form of public-interest digital infrastructure. A case study centered on Puerto Rico illustrates how contextually embedded AI systems can support multilingual communities, local institutions, and public-facing services while preserving privacy and data sovereignty. This record corresponds to a preprint intended for dissemination via SSRN and later arXiv and serves as an archival reference for the Local AI framework and its foundational terminology.
Yann Davidson Vossah (Mon,) studied this question.