August 2025 Summaries

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.

Agentic AI, a rapidly evolving autonomous system capable of independent decision-making and learning, is projected to be incorporated into 33% of enterprise software applications by 2028. This shift presents significant data governance challenges distinct from those faced by traditional AI, necessitating a reevaluation of data security and privacy strategies. Agentic AI's ability to access and process data autonomously raises concerns about data access control, privacy, accountability, and regulatory compliance. Traditional governance models, designed for human-operated systems, may not be equipped to manage the complexities introduced by agentic AI, prompting the need for dynamic access control policies, enhanced transparency, and collaboration across disciplines. Privacy-preserving technologies like homomorphic encryption and federated learning are becoming essential to safeguarding sensitive data, while interdisciplinary efforts and cultural shifts toward responsible AI are crucial for effective governance. Duality Technologies offers solutions leveraging privacy-enhancing technologies to address these challenges, enabling organizations to deploy agentic AI securely and compliantly.