Anthropic’s been eyeing confidential AI. It just made its move with another acquihire.
The New Stack Meredith Shubel ● Covered by 10 sources
Anthropic has hired part of Opaque Systems and licensed its tech. The move gives Anthropic more confidential AI know-how, but leaves Opaque’s open-source future unclear.
Based on reporting by The New Stack, Meredith Shubel — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Anthropic has pulled off another acquihire, this time with Opaque Systems. On Tuesday, Opaque said it struck a new agreement with Anthropic that will “carry Opaque’s work forward.” The wording is soft. The deal is not: Anthropic is licensing Opaque’s technology and bringing over some of its staff, including CEO Aaron Fulkerson and CTO Rishabh Poddar, according to The Information.
What Anthropic gets is access to a niche it has been circling for a while: confidential AI. Opaque says it helped define that category after starting as a UC Berkeley RISELab project 10 years ago, aiming to let companies use sensitive data without exposing it while the data is being used. In plain English, this is about keeping hardware-level protections in place even when the system is actively working on the data. That matters most when the data is too sensitive for the usual “trust us” routine.
Opaque has been doing more than talking. Earlier this year it closed a $24 million Series B at a $300 million valuation. It also built a little ecosystem around its ideas: the Confidential Computing Summit, which Fulkerson says has now been handed to the Linux Foundation; an open standard for verifiable AI governance created with the UAE’s Technology Innovation Institute; AgenTrust, an open-source community for verifiable agent governance; and Confidential MCP, which Opaque describes as the first verifiably governed, secure MCP implementation.
Anthropic, meanwhile, has been signaling interest in this area for a while. Its chief information security officer Jason Clinton appeared on Opaque’s podcast in 2024. In 2025, Anthropic published work with Pattern Labs on confidential inference and called it early, preliminary research. By early 2026, Anthropic showed up on Opaque’s own list of customers and partners. So the deal looks less like a sudden lunge and more like a long, careful reach.
The messy part is what happens next. Neither company has said much about Opaque’s open-source projects, maintenance commitments, or what becomes of the community around them. Some people in the field are shrugging that off. Others see a familiar problem: once a model vendor owns part of the machinery that proves trust, the trust story gets a little less tidy. Developers are being told to pin versions and mirror repositories. That’s usually the kind of advice people give when they’re worried the floor might move.
My take — AI-written commentary, not fact-checked reporting
This is the old open-source trap with a shinier logo: the hard work gets absorbed, the community gets thanked, and the maintenance question gets treated like a minor side quest. Anthropic may be serious about confidential AI, but serious companies also know how to vacuum up a category before the rest of the field has even finished writing the docs.
Read more about this at: The New Stack
Related stories
Inside Key Changes in Data Policies, Ox Alpha Revealed, Taking Custom Models Beyond Fine-Tuning
The Batch ·
49
Inside Ode with Anthropic, the startup betting AI services are the future of enterprise
TechCrunch · 3 months ago ·
51