On-device models are AI components stored and run locally on a computer, phone, or other device instead of being fetched from a remote service for each use. Depending on the system, local processing can reduce the need to send data over a network and may allow some features to work offline. Model files can also occupy substantial storage and may be updated or removed separately from the feature’s visible controls.
In cyber security, local execution changes where data is processed, but does not by itself guarantee privacy or safety: an application may still transmit prompts or results, and vulnerable software can expose local files. In real deployments, defenders should verify which models and related data are installed, how they are updated, and whether disabling an AI feature actually removes its components. Device-management policies, storage checks, and documented settings help confirm the system’s state rather than relying only on a switch in the interface.