Federated Learning (FL), enhanced by Privacy Enhancing Technologies (PETs), is emerging as a powerful solution for training AI models on distributed, sensitive, and regulated data without compromising privacy or compliance. By allowing computation to occur locally and sharing only model updates rather than raw data, FL enables collaboration across sectors such as finance, healthcare, and cross-border research, where data privacy and regulation are paramount. Duality Technologies advances this by integrating PETs like Trusted Execution Environments and Fully Homomorphic Encryption into the FL process, further securing data during aggregation. Additionally, differential privacy and governance mechanisms are employed to protect individual data records and ensure data owners maintain control over data access and usage. This approach has enabled significant advancements, such as fraud detection models trained collaboratively by banks and cancer-detection models developed across hospitals, all while maintaining strict compliance with privacy laws and regulations, thereby demonstrating that privacy can be a foundational element of collaborative innovation rather than a barrier.