A LoRA (Low-Rank Adaptation) adapter is a small set of learned parameters added to a pretrained model to adapt its behavior for a particular task. Instead of updating every parameter in the base model, training adjusts these low-rank components, which can reduce training costs and make adaptations easier to store, distribute, and switch.
In cyber security, an adapter may tailor a model for tasks such as classifying alerts or recommending next steps while the base model remains unchanged. A locally run adapter can help keep data within an organization’s environment, but it is not automatically safe or accurate. Defenders should verify its source and version, test it on representative cases, monitor its outputs, and retain rollback options. If its recommendations feed automated response actions, enforce validation and approval controls rather than treating model output as authorization.