Local inference
Cyberdroid AI models run on the organization’s own hardware.
AI Detection, available now to select organizations · AI Investigation, early access
Can the AI models run on premises?
Yes. Cyberdroid AI Detection and Investigation run their models on premises by default.
See the methodMethod
Inference stays inside the organization’s environment unless the team turns on an online fallback.
Model serving runs within the organization’s environment, with access and workload limits set for the deployment.
The routing record shows which model served each request.
Detection runs on CPUs in the organization’s environment, with no GPU.
Any online fallback is a visible policy the team sets.
Measures
A proof of value measures resource use and response time on the proposed hardware, using the expected workload.
Limits
Model choice and hardware affect capacity, latency and investigation quality. Evaluation checks those tradeoffs on the cases the team needs to handle.
