Deep Learning, a branch of artificial intelligence, could allow an optimal research of legislation and legal doctrine. Imagine a super computer that could know all the intricacies of the law of each country. That's what Deep Learning could accomplish. Deep Law could find solutions to the conflicts that humans could not imagine. Deep Law will use XLNET to understand the language as well as a disentangled variational autoencoder, an LSTM model (Long Short-Term Memory) and a transformer to build a recommender system for lawyers. The objective of this approach is to optimize the work that is preliminarily done by a restricted Boltzmann machine. A training of agents by Reinforcement Learning for legal subjects is also on the agenda. Deep Law will not give legal opinions, but information based on algorithms. Our agent in AI, Deep Law, will be available to support research for lawyers and the general public. Legal algorithmic is a new exciting discipline.