Anthropic’s done it again. On Tuesday, Opaque Systems announced a new agreement with Anthropic that “carries Opaque’s work forward.”
Vague wording aside, the agreement marks another acquihire for Anthropic, with the AI juggernaut licensing Opaque’s technology and bringing on board some (but not all) of its employees, including Opaque CEO Aaron Fulkerson and CTO Rishabh Poddar, reports The Information.
The agreement gives Anthropic access to expertise in confidential AI, the “real-world technology” and category Opaque claims it helped define after starting as a research project at UC Berkeley’s RISELab 10 years ago with the goal to “enabl[e] organizations to use their most sensitive data while keeping it private, even while it’s being used.”
For now, both Anthropic and Opaque are pretty mum on the details of the deal, leaving one obvious question on everyone’s mind:
What will happen to Opaque’s open-source ecosystem? And what does that mean for the developers relying on them?
What is confidential AI? And what does Opaque bring to the table?
On LinkedIn, where he shared news of the agreement, Fulkerson reflected on Opaque’s work defining the category of confidential AI, what he described as “verifiability for the AI stack.”
When asked for more information on the term, Arjun Jaggi, associate director of R&D innovation at Cognizant, describes confidential computing in AI to The New Stack as “running workloads on sensitive data while the data stays protected by the hardware itself, even while it’s being used.”
Whether or not it truly defined the category, the startup’s sure been making its mark. Earlier this year, it closed a $24 million Series B with a $300 million valuation. It’s also spearheaded a slew of industry initiatives.
It created the Confidential Computing Summit, an event Fulkerson says it has now handed over to the Linux Foundation, which already co-hosted this year’s June summit with Opaque. Working with the UAE’s Technology Innovation Institute, it created an open standard for verifiable AI governance. In June this year, it launched AgenTrust, an open-source community for verifiable agent governance, and Confidential MCP, which Opaque bills as the first verifiably governed, secure MCP implementation.
Now, as Opaque enters into a deal with one of the biggest names in the AI space, the obvious question is what will happen to the technology and community it helped build. So far, reactions are mixed.
Anthropic’s been transparent about its interest in confidential AI
In 2024, Jason Clinton, Anthropic’s chief information security officer, appeared on Opaque’s AI Confidential podcast, where he discussed the importance of confidential computing for the future of AI. Afterward, Fulkerson recapped the conversation with a post, the title of which made clear Anthropic’s confidential AI aspirations: “The Mathematical Case for Trusted AI: Why Anthropic is All-In on Confidential Computing.”
Then in 2025, Anthropic published research with Pattern Labs on confidential inference, an approach to protecting sensitive data and AI model weights: “Confidential Inference via Trusted Virtual Machines.” The AI company acknowledged the work was still preliminary, calling it “just a sketch of our research” that was “still early.”
And by early 2026, Anthropic was named in Opaque’s list of customers and partners.
“When a frontier lab like Anthropic takes a promising confidential computing tool and puts real resources behind it, that’s often how a research stage project becomes something hardened enough for production — and an under-funded startup might never get it there on its own.”
Looking back, those public statements suggest Anthropic has been thinking about confidential AI well before its Opaque deal. Whether the new agreement will benefit the industry as a whole, though, remains unknown.
Jaggi, at least, is taking an optimistic stance, telling The New Stack Anthropic’s resources could give the technology the push it needs to grow:
“I’d actually frame this more positively than most. When a frontier lab like Anthropic takes a promising confidential computing tool and puts real resources behind it, that’s often how a research stage project becomes something hardened enough for production — and an under-funded startup might never get it there on its own.”
Others are more skeptical, though.
When asked for his take, Michael Heinrich, co-founder and CEO of 0G Labs, tells The New Stack he sees the agreement as a good move for Anthropic, which can piggyback on Opaque’s decade of confidential computing research. What he expresses caution about is 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.”
Both are opaque on the details of the deal
In his LinkedIn post, Fulkerson describes his move to Anthropic as a natural fit, writing, “Their culture, their mission and their values align with mine, personally and professionally.”
In its announcement, Opaque itself is similarly vague, calling the agreement a “next chapter” that “carries Opaque’s work forward.”
Anthropic, for its part, has told The Information that it’s licensing Opaque’s technology and hiring some of its engineers but has otherwise released no further information.
“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.”
So the (unanswered) elephant in the room is: What happens to all of Opaque’s existing initiatives? The governance framework? AgenTrust? Confidential MCP?
We know that Opaque has passed its summit to the Linux Foundation. But questions remain about the startup’s open-source projects, as neither Opaque nor Anthropic has made any statements about maintenance commitments.
Heinrich seems relatively unconcerned for now, telling TNS that Confidential MCP is MIT-licensed, so anyone can continue using it. He also adds, “Anthropic has set a good precedent here: It gave MCP to the Linux Foundation’s Agentic AI Foundation so no single company controls the protocol.”
Nevertheless, both he and Jaggi advise developers to take precautions now by pinning versions of the tools they rely on and mirroring repositories to preserve code access.
For now, it seems clear Anthropic will benefit from Opaque’s expertise—whether the rest of the industry will, too, remains to be seen.
The post Anthropic’s been eyeing confidential AI. It just made its move with another acquihire. appeared first on The New Stack.