What if in a near future Artificial Intelligence (AI) becomes human-centric, focusing on human needs and build trustworthiness by mutual understanding? Today, millions of people worldwide suffer from deteriorated motor abilities, due to stroke, brain tumor surgery or accident. This represents a serious society challenge with missing adequate technological response. Patients need assistive devices that are trustworthy, multifunctional, adaptive and interactive, i.e., intelligent, unlike current neuroprosthetics that replace single motor impairments. MAIA proposes a paradigm shift where human-centric AI will control prosthetic and assistive devices. We will investigate and resolve critical steps towards the rapid development of such human-centric control: a radically novel intention decoder, a novel concept for trustworthy human-AI interactions, and a new type of database for acquired information from multiple sources. MAIA AI technology will decode human intentions and communicate the decoded targets to assistive devices and to the users, to ensure compliance and develop trust through natural interaction and mutual learning. The technological outcome will be a multifunctional human-centric AI controller at TRL4 with embedded trustworthy characteristics, suitable to be integrated in robotic arms, wheelchair and exoskeletons. To reach this, MAIA will investigate the principles underlying natural, fast and lean communication and new forms of combinations of neural and behavioural data beyond current data processing. MAIA’s approach will be guided by real needs of end users (patients and caregivers) through their direct involvement in the research program, and by all current knowledge from neuro-, cognitive, and social science research. The application domains of MAIA’s new paradigm span from healthcare to industry, and space exploration. We will also establish a European innovation ecosystem beyond the research labs that will stimulate highly innovative enterprises.