OPAQUE Acquires Cryptographic AI Technology from Technology Innovation Institute (TII) to Expand Confidential AI Capabilities

A new acquisition is bringing advanced cryptographic research into real-world AI deployments, as OPAQUE acquires technology developed by the Technology Innovation Institute (TII). The move reflects a growing effort to make it possible for organizations to use highly sensitive data in AI systems without exposing it.

At the center of the deal is TII’s work in privacy-preserving computation and post-quantum cryptography. These capabilities are now being integrated into OPAQUE’s platform to support AI applications in industries such as healthcare, finance, and government, where strict data protections have limited adoption.

A Meeting of Two Different Strengths

OPAQUE was founded by researchers from UC Berkeley’s RISELab, including figures closely associated with the development of modern data infrastructure and distributed systems. The company has positioned itself around a core idea: that AI cannot scale in enterprise environments unless privacy and governance are built directly into how systems operate. Its platform focuses on “confidential AI,” enabling organizations to run models, workflows, and increasingly AI agents while maintaining strict control over how data is accessed and processed.

TII, on the other hand, represents a different part of the technology ecosystem. As the applied research arm of Abu Dhabi’s Advanced Technology Research Council, it focuses on building foundational technologies across multiple domains, including cryptography, quantum computing, autonomous systems, and AI. Its cryptography division has been particularly active, contributing to advanced techniques that allow computation on encrypted data and exploring protections designed for a future where quantum computing could break current encryption standards.

This acquisition connects those two layers. It takes technology that was developed in a research environment and places it into a platform designed for enterprise deployment at scale.

Extending AI Across the Full Lifecycle

OPAQUE’s platform was already focused on secure AI inference, ensuring that models could operate on sensitive data without exposing it. With the addition of TII’s technology, that protection now extends into earlier stages such as training and fine-tuning.

This is a meaningful shift. Training has traditionally been one of the most difficult stages to secure, because it requires large volumes of raw data. Techniques like multi-party computation and fully homomorphic encryption make it possible to train models without revealing the underlying data, even to the system performing the computation.

At the same time, the inclusion of post-quantum cryptography introduces protections designed to remain secure even as computing power evolves. While quantum computing is still emerging, its potential to break widely used encryption methods has made long-term data protection an increasing concern for governments and large enterprises.

Why Sensitive Data Has Been Difficult to Use

Many organizations already possess data that could significantly improve AI outcomes. This includes patient records, financial transactions, and proprietary research. However, these datasets are often underutilized because the risks associated with using them are too high.

Most existing solutions address only part of the problem. One system might secure training, another might handle deployment, and additional tools are required for governance and compliance. This fragmented approach introduces complexity and leaves gaps that can prevent projects from moving into production.

OPAQUE’s approach is to unify these stages into a single system where data usage is continuously verified. Instead of relying on trust in vendors or infrastructure providers, the system generates cryptographic proof that policies were followed and data remained protected throughout the process.

Built for Regulated and Sovereign Environments

A key aspect of the combined platform is its reliance on hardware-based security, including Trusted Execution Environments. These environments isolate workloads at the hardware level, ensuring that even the platform provider cannot access the data being processed.

This design makes the platform suitable for highly regulated environments and for sovereign AI initiatives, where data must remain within specific geographic or legal boundaries. Governments and enterprises can deploy AI systems while maintaining control over data residency and compliance with regional regulations.

It also opens the door to cross-border collaboration on sensitive datasets, where multiple parties can contribute data without exposing it to one another.

The Direction of AI Infrastructure

The acquisition reflects a broader shift in how AI systems are being developed and deployed. As organizations move beyond experimentation, the ability to use sensitive data safely is becoming a central requirement.

AI systems are also becoming more autonomous, with agents capable of executing tasks, interacting with systems, and making decisions. This increases both their usefulness and their risk, particularly when they are connected to critical infrastructure or regulated data.

In this context, technologies like confidential computing and cryptographic verification are likely to become foundational. The emphasis is shifting toward systems that can continuously prove they are operating within defined constraints, rather than relying on assumptions about security.

By combining advanced cryptographic research with a production-ready platform, this acquisition highlights how the next phase of AI adoption may be shaped less by new models and more by the infrastructure that makes them safe to use at scale.