The issue isn’t even that they copy it once anymore. That would be a one and done thing.
The biggest problem is that they can’t even be arsed to copy anything but the URL. They save their own resources by feeding only the address and metadata into the training model and then letting it loose to collect the datasets for itself again.
That means every new model and every slight tweak is an additional crawler that will spamm your server with requests, all because these lunatics are too cheap to buy their own hdds.
The issue isn’t even that they copy it once anymore. That would be a one and done thing.
The biggest problem is that they can’t even be arsed to copy anything but the URL. They save their own resources by feeding only the address and metadata into the training model and then letting it loose to collect the datasets for itself again.
That means every new model and every slight tweak is an additional crawler that will spamm your server with requests, all because these lunatics are too cheap to buy their own hdds.