The deployment of AI in life sciences is a matter of profound public trust. We are handling the most sensitive data imaginable, from a personal genome to confidential clinical trial results. A failure here can lead to patient harm, catastrophic regulatory penalties, and a permanent erosion of public confidence. This chapter argues that trust must be engineered into every AI system from its inception. We first propose four pillars: privacy, confidentiality, security, and compliance that trust can be built on. We then move from abstract principles to concrete technical solutions, exploring the arsenal of Privacy-Enhancing Technologies (PETs) like Federated Learning and Confidential Computing that enable collaboration without compromising sensitive data. We also confront the unique threats against AI itself, from data poisoning to adversarial attacks, and outline a defense-in-depth strategy to safeguard these powerful models.