The base models that can be used with Simon to augment or replace training are built from other peoples speech samples. In order to create high quality base models, a large amount of training samples are necessary.
If you trained your local Simon installation, you gathered valuable voice samples that could improve the quality of the general model.
Through Simon's "Contribute Samples" dialog you can upload those recordings to benefit the Voxforge project to create high quality open source base models.
After connecting to the server, Simon will ask for some basic meta-information. This information obviously contains no personal information. Instead, it will later be used to group together samples of similar speaker groups to build more accurate acoustic models.
The duration of the upload process itself will depend on your internet connection. Generally speaking, this only transmits relatively little data because the audio samples collected by Simon are generally very small: around 0.1 MB per sample.