Towards Trustworthy Ai Systems: Transparency, Interactivity, And Adversarial Assessment