Dear reader,

We're kicking 2025 off at Opaque with big news and ambitions to drive human progress by empowering organizations to harness AI responsibly by maintaining verifiable data privacy and sovereignty. Over the past few weeks, we've hit several exciting milestones that reflect the momentum and innovation we're driving forward—and we're just getting started.

One of the biggest updates is on the product front. We’re introducing our new Compound AI for Agents architecture, enabling everything from AI workflows and pipelines to agentic AI systems with verifiable privacy guarantees. This advancement positions Opaque as the only confidential computing platform that enables organizations to deploy autonomous AI agents while maintaining complete control over sensitive data and models through cryptographic attestation and verification.

These updates unlock new possibilities for secure AI-driven applications that can interact with real-world systems and data while preserving data privacy and model security. Read more in this issue's featured article.

We're also celebrating some big news from our co-founder, Raluca Ada Popa, who is joining Google DeepMind as a senior staff research scientist. Raluca's work on security and privacy has shaped the foundation of Opaque, earned her a Grace Hopper Award, and now she's taking on new challenges in generative AI safety and frontier security at DeepMind. While she steps into this exciting position, Raluca remains a champion of Opaque's mission. Besides being a co-founder, she is also on the Opaque board of directors. Raluca remains a professor at UC Berkeley and a Co-Director of the Skylab. We're incredibly proud of her ongoing contributions to the field and we’re eager to learn from her research with DeepMind. 

Finally, we're already gearing up for the Confidential Computing Summit in-person in San Francisco on June 17-18, 2025. After a successful event last year with speakers and announcements from NVIDIA, Azure, and Google, we've established CC Summit as the year's event for building confidential data, AI, and computing. We intend to make 2025's summit even better—if you have valuable insights to contribute, we encourage you to submit a speaker proposal. Register now to secure your spot with an "Early Bird" discounted rate, and stay tuned for more updates!

Here's to a year of innovation, collaboration, and pushing boundaries responsibly with confidential AI.

Aaron Fulkerson
CEO, Opaque Systems

Expanding the Possibilities of Confidential AI: Introducing Compound AI for Agents

The AI landscape is rapidly evolving, and two major trends are reshaping how enterprises approach artificial intelligence. First, there's a growing shift toward multi-component AI systems that significantly outperform traditional monolithic models. Second, organizations are demanding and needing stronger data privacy controls than ever before. This creates a critical challenge: how can companies deploy sophisticated, compound AI systems—particularly autonomous AI agents—that can safely work with sensitive data while maintaining robust security?

Until now, organizations have been forced to choose between capability and confidentiality, limiting their ability to extract value from their most sensitive data and models. Opaque's latest breakthrough changes this paradigm with our Compound AI for Agents architecture. This innovative solution enables secure, autonomous compound AI systems with cryptographic privacy guarantees and complete auditability—allowing organizations to accelerate their AI adoption while maintaining trust and compliance.

We're witnessing a fundamental transformation in how enterprises use AI. Rather than relying on isolated models, organizations are embracing what we call Agentic Workflows—dynamic systems of specialized AI agents that can reason, plan, and take action. These workflows range from straightforward analysis pipelines to sophisticated multi-agent systems capable of handling complex, multi-step tasks autonomously.

Our Compound AI for Agents architecture was specifically designed for this new era. It ensures that every aspect of agent reasoning and tool usage maintains verifiable privacy and security. The platform offers remote attestation to verify agent authenticity, confidential computing environments to secure agent reasoning and tool usage, comprehensive audit trails to verify agent actions, and robust protection for valuable models—especially those enhanced through targeted learning.

Building on our core "Services" primitive, we're developing a comprehensive suite of capabilities. These include serving of small and large language models, agentic tools like extensions, functions, and data stores, orchestration layers for autonomous operation, and advanced real-time data processing features for cleansing, conformity, and transformation.

For organizations that need to leverage sensitive data for AI operations while maintaining the highest standards of privacy and trust, our platform offers an unmatched combination of capability and security. It's particularly valuable for enterprises looking to build compound AI systems that can derive deeper, more valuable insights while ensuring complete data protection and compliance.

This breakthrough represents more than just technological advancement—it's about enabling organizations to fully realize the potential of AI while maintaining complete control over their sensitive data and operations. As the AI landscape continues to evolve, we're committed to helping enterprises stay at the forefront of innovation without compromising on security or privacy.

Ready to see how Compound AI for Agents can transform your AI initiatives? Schedule a demo today.

In the Lab

The Latest Happenings at Opaque Systems

Call for Speakers: Confidential Computing Summit 
Is your organization leveraging confidential computing or privacy-preserving AI to protect sensitive data? We invite you to share your insights, challenges, and success stories with a community of experts and industry peers at the Confidential Computing Summit, June 17-18, 2025, in San Francisco. Whether you’ve developed new methods to safeguard data or are advancing what's possible in the generative AI era, we encourage you to submit a proposal by Jan. 17. The Confidential Computing Conference Program Committee will review all submissions and select the most compelling presentations for the event.

Beyond Microservices: How AI Agents Are Transforming Enterprise Architecture
Consider how a typical microservice operates: it receives a request, processes it according to predetermined business logic, and returns a response. Now, imagine replacing that rigid service with an intelligent agent that perceives its environment, makes autonomous decisions, and takes purposeful actions to achieve specific goals. This shift represents a profound transformation in how systems operate. For a deeper dive into the potential and mechanics of this concept, explore Google's whitepaper detailing AI agents, along with Aaron Fulkerson’s full blog post.

Machine Minds Podcast: The Unlocking of Confidential Artificial Intelligence with Chester Leung
Episode 75 of the Machine Minds podcast with Greg Toroosian features Chester Leung, co-founder and head of platform architecture at Opaque. Chester shares his journey from a non-technical college student who stumbled into computer science at UC Berkeley to his involvement in cutting-edge data privacy research. The conversation explores what confidential computing actually is—and how it allows organizations to analyze and derive insights from sensitive data without exposing it. Chester also explains how Opaque’s platform empowers data scientists and engineers to process encrypted data and run machine learning workloads efficiently, all while preserving user privacy and meeting stringent compliance demands.

Securing the AI Renaissance: Reflections from the Engine Room 
The transformative potential of confidential computing for AI parallels the early days of the internet, where security protocols enabled trust and scalability. By addressing the compounded risks of data exposure in AI systems, Opaque provides safe environments for AI agents to process sensitive data while maintaining privacy. This foundational work builds the trust necessary for organizations to confidently scale AI operations and unlock real-world impact. Read Aaron Fulkerson’s full blog post. 

AI Confidential Salon: Hosted by Opaque
Join us in San Francisco on Feb. 18 for an intimate, salon-style gathering of technical leaders exploring how to drive AI innovation without sacrificing data security or control. Opaque is bringing together Raluca Ada Popa and Aaron Fulkerson to connect with other forward-looking leaders in a friendly, conversation-oriented environment—no stale lectures, no sales pitch. Request an invitation today to join us for an evening of shared knowledge, cutting-edge AI insights, and authentic peer connections.

Community Corner

Insights from AI Confidential Readers

Jigyasa Grover on How Opaque Empowers Financial Institutions to Unlock the Full Potential of Collaborative AI
Data fuels the finance industry, but ensuring data privacy and sovereignty across global operations is a growing challenge. Opaque empowers financial institutions to unlock the full potential of collaborative AI while keeping sensitive data protected and compliant. This isn't just about avoiding risk; it's about gaining a competitive edge. Read Jigyasa’s full post.

Sanchit Shangari on How Confidential AI is Here to Turn Responsibility into Limitless Opportunity
With great power comes great responsibility—and Confidential AI is here to turn responsibility into limitless opportunity. AI is reshaping everything from predictive analytics and fraud detection to multi-party data sharing and supply chain optimization, with Confidential AI enabling businesses to safely use sensitive data, unlock value from collaboration, and future-proof their operations. More than a technology, Confidential AI is a trusted, foundational platform for innovation in an era defined by data-driven decision-making. Read Sanchit’s full post.  

Jigyasa Grover on How Opaque Powers the Insurance Industry with Confidential AI
The insurance industry thrives on your data, but sharing too much can expose sensitive information. Opaque helps insurers maintain complete data privacy and sovereignty while collaborating securely on AI-powered initiatives. Read Jigyasa’s full post.

Community Corner is a space for you to share the posts, articles, podcasts, or videos that have sparked your curiosity or shifted your perspective on AI and innovation. Tell us what you’re reading, share a conversation you’re following, or send us your favorite videos and podcasts. 

Click below to submit your recommendation, and we’ll feature select contributions in future issues of the newsletter.

Code for Thought

Worthwhile Reads

🔜 The way AI is built is about to change. OpenAI co-founder Ilya Sutskever said that while he believes existing data can still take AI development farther, the industry is tapping out on new data to train on, forcing a shift away from the way models are trained today. The next generation of AI will involve agents leveraging highly sensitive enterprise data, making Confidential AI solutions critical to ensuring data privacy and compliance in advanced applications. 

🔑 In 2025, key factors will influence the cybersecurity market. IT spending is projected to grow by 10% in 2025, driven by a 35-40% surge in AI investments and a 15% rise in cybersecurity, with security comprising up to 20% of IT budgets. As organizations prioritize identity management, compliance, and advanced threat mitigation in 2025, integrating confidential computing solutions will enhance privacy functions while safeguarding sensitive data. 

🚗 Data is driving innovation in today’s finance industry. GenAI is revolutionizing the finance industry, enabling financial institutions to extract actionable insights, enhance risk management, and deliver hyper-personalized customer experiences. As AI adoption accelerates, financial leaders must act strategically to integrate these technologies, leveraging their transformative potential to innovate, compete, and lead the future of finance.

🧠 Humans often use tools to supplement their knowledge, and GenAI models can do the same. But how? By accessing real-time information or performing real-world actions. For instance, AI can use tools like databases for tailored recommendations or APIs to automate tasks like sending emails or completing transactions. This ability to reason, plan, and integrate external tools defines an "agent," extending AI's capabilities beyond standalone models—a concept explored in detail in this whitepaper.