Locally hosted AI is an artificial intelligence system run on infrastructure controlled by the operator, instead of being accessed through a third-party cloud service. Models, prompts, logs, and generated outputs stay inside that environment, which gives the operator more control over data handling, latency, and configuration.
In cyber security, that matters because local hosting can reduce outside visibility. Defenders may have less insight into what prompts were used, what documents were processed, or how the model was tuned. Attackers can use that privacy to support phishing, translation, malware development, document triage, and task automation while keeping sensitive material off external services. For defenders, the term is relevant because it highlights a workflow risk: the AI itself is not the threat, but it can make familiar intrusion steps faster, more scalable, and harder to observe. Monitoring email abuse, identity controls, and post-click behavior remains essential.