Upper limb amputation is a major injury that can strongly affect the daily life of a person. Prosthetic hands that can execute multiple movements are available, but they are expensive and difficult to control. Natural control via pattern recognition is promising but it is applied in real life prosthetics only in limited ways. Additive manufacturing and machine learning can revolutionize prosthetics with affordable and open-source solutions that can include 3D printed prosthetic hands, sockets and dexterous highly functional control. Nevertheless there are still intermediate steps to do into this direction. The objective of this paper is to introduce an ongoing project aimed at the development of low cost and dexterous prosthetic hands to be used in real life conditions based on open 3D models, additive manufacturing and machine learning. The results at the current state of advancement of the project include several versions of the prosthetic hand (powered by six servomotors and based on open design), of the control system (based on open electronic prototyping platforms) and of the socket. Preliminary tests of the hand demonstrate its dexterity, its potential and requirements to improve force. Once fully completed and released, the presented3D printed, dexterous, open-source prosthetic hand has the potential to improve the life of hand amputees worldwide and to foster improvements in research and for future commercial prosthetic hands.