Towards Assuring EU AI Act Compliance and Adversarial Robustness of LLMs