What does local mean?
With local AI, the model computation runs on hardware you operate or control. That can be one computer, a local server or your own technical environment. Local does not automatically mean offline, secure or better; those properties need separate evaluation.
What can be useful about it?
Data can remain inside the controlled environment, some systems can work without a permanent connection to an external AI service, and technical boundaries can be managed more directly. At the same time, speed and model size depend more strongly on available hardware.
What belongs in the decision?
Data protection and access are only part of the picture. Maintenance, updates, model quality, hardware and power requirements, and the security of the local system matter too. Some tasks fit cloud services better, while others benefit from local or hybrid architectures.