June 2026 Summaries

Navigating GDPR compliance in cross-border AI data collaboration is complex, as it requires understanding specific rules, penalties, and the limitations of anonymization, especially when involving sensitive EU personal data across various sectors. GDPR's extraterritorial reach means that any AI systems processing data from EU residents fall under its scope regardless of the organization's location, emphasizing the importance of Article 25, which mandates technical controls beyond mere policy documentation. Privacy-Enhancing Technologies (PETs) like Fully Homomorphic Encryption and Federated Learning are critical in supporting data protection by design, enabling secure AI collaboration without compromising data privacy. Legal mechanisms such as Standard Contractual Clauses and Transfer Impact Assessments are essential for ensuring data protection during cross-border transfers, which can be further complicated by the need for Binding Corporate Rules in multinational environments. Duality Technologies exemplifies how integrated solutions can fulfill GDPR obligations, allowing for secure and efficient AI collaboration without exposing raw data, demonstrating that compliance and innovative AI solutions can coexist effectively.

Duality Technologies and Red Hat have formed a partnership to enhance secure analytics and AI without exposing sensitive data, particularly benefiting governments and regulated enterprises. This collaboration combines Red Hat's open hybrid cloud and confidential computing technologies with Duality's privacy-enhancing solutions, enabling organizations to operationalize sovereign AI across hybrid and sensitive data environments without centralizing or exposing data. The partnership aims to support secure large language model training, cross-domain analytics, federated AI collaboration, and privacy-preserving analytics, making it especially relevant for sectors like defense, public sector, healthcare, and financial services that need to comply with strict privacy and security requirements. Dr. Alon Kaufman of Duality Technologies highlights the importance of this collaboration in unlocking modern AI capabilities for organizations handling sensitive data, while Axel Sass from Red Hat emphasizes the need for secure infrastructure and collaboration to deploy trusted AI effectively.

The collaboration between Red Hat and Duality Technologies introduces a comprehensive confidential computing architecture designed to protect data in use, addressing a critical gap in data security. This platform leverages Red Hat's OpenShift confidential containers and Duality's advanced privacy-preserving technologies, such as fully homomorphic encryption and federated learning, to enable secure collaboration across various environments. Duality's platform offers end-to-end control over data and execution lifecycles, facilitating secure AI operations and data management while ensuring compliance through robust governance, authentication, and auditing features. It supports protected model inference in the cloud by encrypting workloads at the client-side and ensuring they are only decrypted within secure environments, thereby safeguarding against unauthorized access and data breaches. The system's architecture is built upon three layers—collaboration management, computation, and privacy technologies—enabling seamless and secure execution of complex analytics and machine learning tasks.

Data masking and encryption are both essential data protection methods, each addressing different aspects of data security. Data masking replaces real values with fictitious ones, making it irreversible and particularly useful for non-production environments like testing and development, while encryption scrambles data to make it unreadable without a decryption key, ideal for securing data at rest and in transit. However, both methods fall short in protecting data during active processing, exposing vulnerabilities when data is computed on or shared for collaborative analysis. Privacy-enhancing technologies (PETs), such as Fully Homomorphic Encryption (FHE), offer a solution by enabling computation on encrypted data without exposing it, crucial for modern data use cases involving AI and cross-organizational collaboration. Duality Technologies specializes in operationalizing these PETs, allowing organizations to securely process sensitive data without the risk of exposure, thus closing the security gap that traditional masking and encryption leave open during active processing.

Duality 4.6 addresses challenges of analyzing distributed data by introducing Data Harmonization and expanded support for Custom Federated Workloads, allowing organizations to conduct federated computations across diverse datasets without restructuring source systems or moving sensitive records. The new Schema Harmonization Tool enables users to map source columns to a common schema, normalize values, and apply harmonization rules at execution time, facilitating a unified analytical view while maintaining data control. Additionally, the platform enhances flexibility by allowing users to independently develop and deploy federated workloads using existing Python-based code, minimizing the need for code refactoring and reducing operational overhead. This approach ensures that organizations can collaborate effectively across fragmented data environments, preserving local governance and existing workflows while participating in shared analytics and AI initiatives. With added features like Parquet File Support, Edge Filtering, and Index Query Optimization, Duality 4.6 simplifies the deployment and scalability of federated analytics, making it easier to integrate into real-world data environments and extract value from distributed data.

Privacy-preserving computing technologies, particularly Fully Homomorphic Encryption (FHE), have historically faced performance challenges, but recent advancements in cryptography, hardware acceleration, and open-source software have transformed these technologies from research concepts into practical enterprise solutions. Initiatives like DARPA's DEPRIVE program have played a pivotal role in overcoming the performance barriers of FHE, facilitating the development of optimized software libraries and dedicated hardware that can handle encrypted computation efficiently. As organizations recognize that their competitive edge lies in accessing data previously out of reach, Privacy-Enhancing Technologies (PETs) are becoming essential for secure AI infrastructure, enabling data access without compromising privacy or security. These technologies, including FHE, federated learning, Trusted Execution Environments, and differential privacy, allow for secure collaboration across decentralized datasets and support AI systems that span multiple jurisdictions and organizations. The shift towards PETs signifies a move from focusing solely on model performance to emphasizing data access and trust, thereby integrating these technologies into enterprise AI strategies as enablers of business opportunities while maintaining compliance and security.

Data governance and data architecture are often misunderstood in enterprise data strategy, with many organizations focusing on one while neglecting the other. Data governance involves setting policies, access controls, and accountability measures to determine who can access what data and for what purpose, while data architecture is about the technical design that ensures data is stored, moved, and processed efficiently. Successful organizations integrate both, ensuring that governance defines what to enforce and architecture determines how to enforce it, creating a seamless operational layer where policy enforcement is embedded in system design. Legacy systems present significant challenges for governance frameworks, often requiring costly and complex integrations to achieve compliance. A data governance architect plays a crucial role in bridging the gap between policy and technical enforcement, ensuring that governance intent is realized in production environments. Both governance and architecture must evolve together to support regulatory compliance and secure data sharing, particularly when data crosses organizational boundaries.

Sovereign cloud refers to a cloud computing environment designed to align data, infrastructure, and operational controls with a specific jurisdiction and governance framework, aiming to reduce exposure to foreign legal access and unauthorized intrusions. It has become crucial for governments and regulated sectors, driven by stricter data protection requirements and geopolitical pressures. The concept of sovereignty in the cloud extends beyond data residency to include control over data, infrastructure, and encryption, challenging traditional public cloud models that often fail to provide full jurisdictional independence. The CLOUD Act complicates data sovereignty, as it allows US authorities to access data stored globally by US-based companies, raising concerns for organizations under regulations like GDPR. True sovereignty requires cryptographic controls that prevent providers from accessing data, employing privacy-enhancing technologies like fully homomorphic encryption and secure multi-party computation, which enable secure cross-border collaboration without exposing sensitive data. However, sovereign cloud solutions often come with increased costs and complexity, and the term itself lacks a universal certification standard, making technical evidence of cryptographic controls essential for validating claims of sovereignty.

Sir Alex Younger, the former Chief of the UK's Secret Intelligence Service and a member of Duality's UK Advisory Board, has passed away, leaving a significant impact on those who knew him. Renowned for his quiet patriotism and unwavering belief in Western values, Sir Alex was not only a remarkable individual but also a source of wisdom and a valued colleague at Duality. His contributions and support to the company's mission were greatly appreciated, and his passing is deeply mourned by those who had the privilege to work with and learn from him.