Artificial intelligence is rapidly evolving from task-specific tools into autonomous systems that increasingly affect both information spaces and the physical world. This talk examines how to build explainable, trustworthy, and ethical AI by combining symbolic and sub-symbolic approaches. It highlights limitations in factual grounding, logical reliability, and faithful explanation, together with challenges of responsibility, fairness, compliance, and value conflicts. Four principles are proposed: factual grounding through world models, controlled reasoning with structured formalisms, verified trustworthiness through explanation, and ethical alignment by design and causality. Computational argumentation is presented as a key approach toward AI that is Correct, Clear, and Compliant.