The dataset is secured, no problems with that, but someone could take the model and use it as it is. Our model is trained and runs on a popular open source framework, which we advertise as a feature since many people are familiar with how well it works already. Our main "product" is the model itself, made by painstakingly labeling hundreds of thousands of videos manually. Unfortunately deep learning models are not considered algorithms here and cannot be patented at the moment. So all we can do is hide it.