As far as I remember OpenAI does it too even when using the API. Their reason is fraud and harmful behaviour detection. But let's be honest, does it really matter? Building a successful product does depend on so much more than the technical implementation and brainstorming you do with Fable, Mythos or any model.