Anthropic Acquihires Opaque Systems, Licensing Confidential AI Technology

Anthropic Licenses Opaque's Confidential AI Technology

Opaque Systems announced on Tuesday a new agreement with Anthropic that the startup described as a "next chapter" that "carries Opaque's work forward." The deal amounts to an acquihire: Anthropic is licensing Opaque's technology and bringing on some, but not all, of its employees, including Opaque CEO Aaron Fulkerson and CTO Rishabh Poddar, according to The Information.

Opaque began as a research project at UC Berkeley's RISELab 10 years ago with the goal of enabling organizations to use their most sensitive data while keeping it private, even while it is being used. Fulkerson has described the company's focus as confidential AI, or "verifiability for the AI stack." Arjun Jaggi, associate director of R&D innovation at Cognizant, describes confidential computing in AI as running workloads on sensitive data while the data stays protected by the hardware itself, even while it is being used.

A Decade of Confidential Computing Work

Opaque has built a track record in the space. Earlier this year it closed a $24 million Series B round at a $300 million valuation. It created the Confidential Computing Summit, which it has handed over to the Linux Foundation, and worked with the UAE's Technology Innovation Institute to create an open standard for verifiable AI governance. In June it launched AgenTrust, an open-source community for verifiable agent governance, and Confidential MCP, billed as the first verifiably governed, secure MCP implementation.

Anthropic's Longstanding Interest

Anthropic's interest in confidential AI is not new. In 2024, the company's chief information security officer, Jason Clinton, appeared on Opaque's AI Confidential podcast to discuss the importance of confidential computing for the future of AI. In 2025, Anthropic published research with Pattern Labs on confidential inference, an approach to protecting sensitive data and AI model weights, though it called the work preliminary. By early 2026, Anthropic was named in Opaque's list of customers and partners.

Mixed Reactions and Open Questions

Reactions to the deal are mixed. Jaggi is optimistic, saying Anthropic's resources could give the technology the push it needs to move from research to production. Michael Heinrich, co-founder and CEO of 0G Labs, sees the agreement as a good move for Anthropic but warns of a potential conflict of interest: "The whole point of confidential computing is that you don't have to trust the operator. When the tools that produce that proof sit inside a model vendor, the party being checked now owns part of the checking."

Neither company has said what will happen to Opaque's existing initiatives, including AgenTrust and Confidential MCP, or whether it will maintain them. Heinrich notes that Confidential MCP is MIT-licensed, so anyone can continue using it, and points out that Anthropic previously gave MCP to the Linux Foundation's Agentic AI Foundation. Both he and Jaggi advise developers to pin versions of the tools they rely on and mirror repositories to preserve code access.

Sources: The New Stack