Anthropic has licensed technology from Opaque Systems and hired some of its staff, an Anthropic spokesperson told The Information on October 13. The spokesperson would not give terms or say how many engineers are moving. Opaque is a five-year-old startup that grew out of UC Berkeley research. It sells software for running AI agents on confidential data.
Opaque raised a $24 million Series B in February at a $300 million valuation, per Constellation Research. Walden Catalyst and Intel Capital were among its investors, per VC News Daily. A license-and-hire pays for rights and people, and nobody has disclosed the price. Holders who bought in at $300 million eight months ago cannot yet tell what their stake returns.
Opaque's own post describes "a new agreement with Anthropic that carries OPAQUE's work forward," per the company. It does not call the deal an acquisition. It says nothing about the product, customer contracts, or support. It does list open-source projects the team worked on, including Confidential MCP, a set of controls on the tools agents call.
This is the structure Nvidia used with Groq. That deal now faces a Justice Department review and a stockholder suit, as we reported. It moves people and rights without buying the company. For Anthropic, Opaque adds a credential that enterprise and government buyers ask for when their data has to stay sealed, including on their own servers.
Bottom Line
Anthropic got Opaque's people and a license, and nobody has said who supports the product now. If you run Opaque's platform, ask for a support and exit timeline in writing this week, and plan as if it ends. If you are a founder in AI security, license-and-hire is now a common way out to a big lab. Model what it pays common shareholders before a term sheet arrives.