5 key questions to ask when developing trustworthy AI models