Opaque deepens trusted AI push

OPAQUE has acquired cryptographic AI technologies developed in Abu Dhabi, expanding its confidential AI platform into model training, agentic workflows and post-quantum protection for enterprises handling highly sensitive data.

The deal links OPAQUE, a San Francisco-headquartered confidential AI company founded by researchers from UC Berkeley’s RISELab, with the Technology Innovation Institute, the applied research arm of Abu Dhabi’s Advanced Technology Research Council. The transaction, announced on Monday, gives OPAQUE technology designed to protect data while it is being used by AI systems, not merely when it is stored or transmitted.

The acquired capabilities focus on two areas that have become central to enterprise AI deployment: confidential model training using cryptographic techniques such as multi-party computation and fully homomorphic encryption, and defences intended to withstand future quantum-enabled attacks on today’s encryption systems. These tools are aimed at banks, healthcare groups, life-sciences companies, insurers, public-sector bodies and defence-linked organisations that want to use proprietary or regulated data without exposing it to unauthorised access.

OPAQUE said the technology would extend its platform across the full AI lifecycle, covering training, fine-tuning, inference and AI agents. That is a significant step because many companies have so far limited generative AI use to low-risk pilots or public data, fearing breaches, compliance failures or loss of control once AI systems interact with operational databases.

AI agents have intensified those concerns. Unlike chatbots that respond to prompts, agents can take sequential actions, call tools, query systems and make decisions at speed. When connected to financial claims, patient records, clinical research, identity systems or classified information, the same autonomy that makes them useful also increases the potential damage from misuse, error or compromise.

OPAQUE’s pitch is that hardware-enforced execution environments and verifiable rules can provide evidence of what ran, where it ran and which controls were applied. That approach is part of the broader confidential computing market, where secure hardware enclaves are used to isolate workloads from cloud administrators, attackers and other software running on the same infrastructure.

The Abu Dhabi-developed technology adds a cryptographic layer to that model. Multi-party computation allows different parties to jointly analyse data without revealing their individual inputs. Fully homomorphic encryption allows computation on encrypted information without first decrypting it. These methods are computationally demanding, but advances in hardware, software optimisation and AI infrastructure have made them increasingly relevant for regulated sectors.

Post-quantum protection adds another dimension. Cybersecurity agencies and standards bodies have warned that encrypted data stolen today could be stored and decrypted later if large-scale quantum computers become capable of breaking widely used public-key systems. The first wave of post-quantum standards has already pushed governments and major technology companies to begin migration planning, especially for data with long-term sensitivity.

For Abu Dhabi, the sale gives international validation to a research strategy built around artificial intelligence, cryptography, quantum technologies and advanced computing. TII has been positioned as a core institution in the emirate’s effort to become a global technology centre, with research programmes spanning cryptography, secure systems, autonomous robotics, quantum engineering and digital science.

H. E. Faisal Al Bannai, Adviser to the UAE President and Secretary General of ATRC, said the transaction showed that technology developed in Abu Dhabi could be adopted globally. He described the acquisition as evidence that the UAE was producing foundational technology rather than merely adopting tools created elsewhere.

Ion Stoica, OPAQUE co-founder and board member, said the future of AI depended on unlocking data that organisations had been unable to use because of security and regulatory constraints. He said the acquired technology would help provide cryptographic evidence across training, fine-tuning, inference and agents.

Aaron Fulkerson, OPAQUE’s chief executive, framed the acquisition around the risks of agentic AI. He said agents operating on sensitive systems could deliver extraordinary productivity, but also cause severe harm if deployed without provable controls. His argument reflects a wider enterprise shift from experimentation to governed production use, where boards and compliance teams are demanding auditability before approving AI systems for core operations.