Establishing and Evaluating Trustworthy AI: Overview and Research Challenges