The Shadow DEX-EE Series is a robust 3-fingered dexterous robot hand developed in collaboration with Google DeepMind for demanding machine learning tasks. Two models are available: the symmetric DEX-EE and the DEX-EE Chiral with human-like kinematics featuring an offset third finger like a human thumb for easier imitation learning and bi-manual tasks, available in left, right, and bi-manual pairs. Both weigh 4.1 kg, stand 350 mm tall, and offer 12 degrees of freedom. High-bandwidth torque and position control loops provide delicate fingertip dexterity. Stereo camera-based optical fingertip tactile sensors with hundreds of taxels per finger and a massive dynamic range provide unprecedented 3D interaction detail, supplemented by multi-taxel 3-DOF tactile sensors on middle and proximal phalanges. Torque and inertial measurement throughout makes the hand sensitive to environmental interactions. Designed for long-running reinforcement learning experiments with high mean time to failure, reduced repair time, resistance to repeated impacts, and graceful shutdown. Fully ROS integrated. Previously used by OpenAI for single-handed Rubik’s cube solving via reinforcement learning, by Google Brain for multi-object manipulation with just 4 hours of real-world data, and by the Human Brain Project. Targets AI, machine learning, and dexterous manipulation research.