May 2026 Summaries

Fully Homomorphic Encryption (FHE) represents a significant advancement in data privacy by enabling computations on encrypted data without decryption, which is crucial in an era marked by data breaches and stringent privacy regulations. Despite its potential, FHE's complexity has historically posed a barrier due to the need for specialized cryptographic expertise and familiarity with libraries like OpenFHE. Duality has addressed this challenge by integrating Claude Code with their FHE Domain Skills, allowing developers to generate professional-grade OpenFHE code using natural language. This system enhances AI capabilities beyond simple prompts, embedding Duality’s expertise into agentic skills that automate complex cryptographic tasks and ensure the code meets privacy and security standards. The application of this technology is demonstrated through use cases like secure neural networks and privacy-preserving data analyses in sectors such as healthcare and finance, highlighting its transformative potential. By simplifying the development of secure computation workflows, Duality is paving the way for a privacy-by-design digital economy where sensitive data can be utilized without exposure.

Data in use protection is a critical cybersecurity measure that safeguards sensitive information during active processing, which traditional encryption methods for data at rest and in transit do not cover. As organizations increasingly rely on AI, cloud infrastructure, and cross-organizational collaboration, protecting data during computation becomes crucial, particularly in fields like healthcare, financial services, and government. Technologies such as Trusted Execution Environments (TEEs), Fully Homomorphic Encryption (FHE), and Secure Multi-Party Computation (MPC) play vital roles in securing data without exposing it during processing or sharing. These methods allow for secure AI training and analytics while maintaining compliance with privacy regulations and enabling collaborative efforts across various domains. Effective data in use protection often involves a combination of technologies and strong governance to ensure compliance, performance, and secure collaboration, as demonstrated by platforms like Duality, which integrate these capabilities to enhance secure AI and analytics workflows.

Data security is fundamentally about protecting data across three states: at rest, in transit, and in use, each with distinct encryption needs. Data at rest, stored in databases or file systems, is protected by AES-256 encryption, ensuring its unreadability without the correct key. Data in transit, which moves across networks, is secured by TLS 1.3, establishing an encrypted channel between endpoints but leaving data unprotected once it reaches its destination. The least protected state, data in use, involves active processing and requires advanced methods like Trusted Execution Environments (TEEs), Fully Homomorphic Encryption (FHE), or Secure Multi-Party Computation (MPC) to ensure data remains encrypted even during processing. Effective data security involves not only cryptographic solutions but also robust governance frameworks that manage who can compute what, on which data, and for what purposes, emphasizing operational oversight to close gaps between encrypted states. Duality Technologies integrates these elements, providing a comprehensive approach to securing data in use, which is crucial for future AI applications that require safe handling of sensitive data.

Fully homomorphic encryption (FHE) has significantly advanced over recent years, evolving from a theoretical concept with impractical performance for enterprise deployment to a viable option for specific workloads. With improvements in algorithms, compiler optimizations, batching techniques, and hardware acceleration, FHE performance has sped up by 1,000x to 10,000x compared to five years ago. This progress has made FHE suitable for batch-oriented tasks such as analytics, machine learning inference, and cross-organization data computations. Despite these advancements, real-time applications with strict latency requirements still pose challenges for FHE, primarily due to bootstrapping overhead, though hybrid architectures can mitigate some limitations by combining FHE with plaintext or trusted execution environments. Hardware acceleration, particularly through GPUs, FPGAs, and emerging ASICs, has been a crucial driver of FHE's growing enterprise viability, with contributions from initiatives like DARPA's DPRIVE program further enhancing performance. While FHE is not yet a universal solution for all real-time systems, its current capabilities offer substantial privacy and performance benefits for asynchronous, data-sensitive, and batch-oriented workloads.

Data at rest refers to any stored data that is not actively moving between systems, and securing it is crucial due to its volume and permanence, which makes it an attractive target for breaches. Unlike data in transit, which is protected by TLS, data at rest requires different security measures, such as AES-256 encryption, robust key management, and access controls. Compliance standards like GDPR, HIPAA, and PCI-DSS mandate encryption at rest, and failing to comply can result in significant penalties. Transparent Data Encryption (TDE) is a key tool for database-level encryption that encrypts data before it is stored and decrypts it when accessed, without altering the application layer. Despite the protection offered by encryption at rest, data becomes vulnerable when decrypted for use, prompting the need for advanced technologies like fully homomorphic encryption (FHE), which allows computation on encrypted data without decryption. Organizations must not only focus on encryption but also on regular audits, appropriate access controls, and effective key management to ensure comprehensive data security.

Large Language Models (LLMs) are transformative tools but pose significant data privacy risks, particularly with sensitive information like personally identifiable information (PII) and proprietary content. These risks manifest at various stages of the AI lifecycle, such as training, fine-tuning, inference, and retrieval-augmented generation (RAG) processes. Privacy-enhancing technologies, including fully homomorphic encryption, federated learning, and differential privacy, are crucial for safeguarding data within LLM deployments, but no single solution is sufficient on its own. Compliance with regulations like GDPR is mandatory, with substantial fines for non-compliance. A robust LLM data privacy strategy involves layered defenses and proactive measures, integrating privacy from the start. Companies like Duality Technologies provide practical solutions by operationalizing these technologies, allowing enterprises to deploy LLMs securely while maintaining data privacy, especially in regulated industries.

The document explores the evolution and challenges of data loss prevention (DLP) policies, emphasizing their need to adapt to modern data environments characterized by AI, cross-border collaboration, and multi-party analytics. Traditional perimeter-based DLP strategies, designed to prevent data movement, often fail in today's fragmented cloud environments and are insufficient against insider threats and AI pipeline vulnerabilities. Modern privacy-first approaches and privacy-enhancing technologies (PETs) are highlighted as crucial for enabling secure and compliant data use without compromising data privacy. These technologies, including homomorphic encryption and secure multi-party computation, allow for data to be analyzed and used while remaining encrypted, thus mitigating risks associated with unauthorized access and regulatory non-compliance. The text underscores the importance of zero-trust architecture in reinforcing DLP strategies, moving enforcement from network layers to data layers, and ensuring that no user or device is inherently trusted. It concludes by advocating for a shift from data blocking to data enablement, where secure computation on sensitive data becomes a priority, allowing organizations to unlock analytical value while maintaining stringent privacy and security safeguards.

Join the inaugural OpenFHE webinar, part of a new monthly series, where the team will present significant advancements, lessons, and future perspectives from DARPA’s DPRIVE program, which focuses on enhancing hardware acceleration for practical Fully Homomorphic Encryption (FHE) and privacy-preserving artificial intelligence (AI).