Researchers at the Technical University of Munich and the Center for Rehabilitation Passauer Wolf have developed a novel soft hand exoskeleton designed to restore motor function in patients with severe neurological impairments. Led by Professor Gordon Cheng, the multidisciplinary team published their findings in Nature Machine Intelligence, detailing a wearable robotic system that integrates muscle-sensing technology, artificial intelligence, and soft robotics. Unlike conventional rigid orthoses that typically require partial voluntary movement, this lightweight device operates as an air-actuated fabric glove. It captures residual electrical activity from the user hand muscles and processes the data through a machine-learning algorithm to predict grasping intentions. Upon detection of a motor intent, the exoskeleton executes precise mechanical movements, enabling users to manipulate objects and perform daily activities. Clinical evaluations included one patient with advanced amyotrophic lateral sclerosis and six individuals recovering from strokes. Participants completed standardized assessments, including the Box and Block Test and the Action Research Arm Test. Results indicated that individuals with the most profound motor deficits derived the greatest benefit. Notably, an ALS patient who had experienced near-complete hand paralysis for four years successfully used the device to feed himself, demonstrating significant functional recovery. Interestingly, patients retaining partial hand control experienced less pronounced improvements, suggesting the system is optimally calibrated for severe impairment cases. A critical finding emerged from the development process: direct patient collaboration proved essential. The ALS participant emphasized the necessity of manual override capabilities, rejecting fully autonomous operation in favor of explicit control mechanisms. This feedback refined the intention-prediction model, ensuring the exoskeleton responds strictly to user-directed neural signals rather than acting independently. The device construction utilizes standard textile manufacturing and pneumatic actuators, prioritizing wearability, safety, and comfort. Researchers note that the integration of intention prediction and soft robotics addresses longstanding limitations in assistive technology, which previously struggled with dexterity and patient autonomy. Looking ahead, the Technical University of Munich team is adapting the underlying architecture for lower-body applications to assist a broader patient demographic. Cheng aims to transition these prototypes into commercial clinical use within the coming years, ultimately expanding access to advanced rehabilitative robotics. The study underscores a shifting paradigm in neurorehabilitation, where AI-driven soft exoskeletons are becoming viable tools for restoring independence to individuals with profound motor disabilities.