Which AI does Herrald use? Is our data used for training?

Anthropic's Claude models. Message text goes to Anthropic only so it can be read and answered, and Anthropic doesn't train on it.

Whatever a message says is always treated as something to read, never as instructions to follow.