From Berkeley Research to the Future of AI: OPAQUE’s Next Chapter

From Berkeley Research to the Future of AI: OPAQUE’s Next Chapter

Ten years ago, OPAQUE began as a research project at UC Berkeley’s RISELab around an idea many thought was impossible: enabling organizations to use their most sensitive data while keeping it private, even while it’s being used.

Together, we turned that research into real-world technology and helped define the category of confidential AI.

OPAQUE brought confidential computing out of the research lab and into production AI environments, making it possible for organizations to run AI workloads on sensitive data with hardware-enforced protections. As AI evolved, so did the problem we set out to solve. Protecting data and models was no longer enough. Organizations also needed a way to independently verify how AI systems and agents behaved.

That led us to expand our work from confidential AI to verifiable AI: replacing trust assumptions with cryptographic evidence of what ran, where it ran, and under what policies.

Along the way, we worked to advance how secure AI systems are built. We developed technology, contributed open source projects and helped move the industry toward a future where security, governance and verifiability are fundamental AI infrastructure.

We also saw that work grow beyond OPAQUE. Through AgenTrust and projects including Agent Governance Toolkit, TRACE, Confidential MCP and Weight Custody Manifest, we worked with a broader community to develop open approaches to some of the hardest trust and security challenges emerging around AI agents and models.

A new chapter

Today, we’re proud to share a new agreement with Anthropic that carries OPAQUE’s work forward. We’re excited to see its impact continue in shaping how AI is built and secured.

For us, this moment is also an opportunity to recognize what the entire OPAQUE team accomplished.

OPAQUE brought together researchers, engineers, product builders, marketers, sellers and operators who took extraordinarily difficult problems and made them practical. We built a company from university research. We helped establish confidential AI as a category. We advanced verifiable AI from an idea into working technology. And we helped demonstrate that some of the hardest security problems in AI could actually be solved.

Most people told us it was impossible. This team proved it was not just possible, but helped set a new standard.

Thank you to every OPAQUEling, and to the customers, partners, investors, researchers and open source contributors who believed in what we were building.

We made the world a little better and safer.

We can all be proud of what we built together.