Anthropic licenses Opaque Systems' confidential AI technology

Anthropic has entered into an agreement to license confidential AI technology from Opaque Systems and has hired several of the startup’s engineers. The arrangement, described by some as an acquihire, is not a full acquisition; Opaque Systems will continue to operate as an independent company.

Opaque Systems, which originated from research at UC Berkeley’s RISELab, specializes in confidential and verifiable AI. Its technology uses hardware-isolated environments and cryptographic evidence to ensure that AI workloads can process sensitive data without exposing it to unauthorized parties, including cloud operators. This allows organizations in regulated industries, such as banking and government, to receive mathematical proof that security policies were followed during computation.

The startup was co-founded by researchers including Raluca Ada Popa and Ion Stoica. Before this deal, Opaque had raised approximately $31.6 million in funding and reached a valuation of roughly $300 million. The company’s technology has been developed in collaboration with major industry players including NVIDIA, Intel, Microsoft, and Google.