Secure Enclave

A secure enclave is a physically isolated region of a processor where code and data can be processed confidentially. Since the isolation is enforced by the hardware itself, as opposed to the operating system or other software, even administrators with the highest level of system access can’t read the data running inside the enclave. 

Also known as trusted execution environments (TEEs), today’s secure enclaves provide stronger guarantees through cryptographic attestation, which verifies that an enclave is genuine, unmodified, and running the expected code before releasing sensitive data or encryption keys to it.  

The isolation-attestation combo makes secure enclaves the foundation of confidential computing for both traditional applications and AI workloads. 

Secure enclaves let teams process sensitive data without exposing it to the underlying infrastructure. The most recent use case to explode is confidential AI inference, where model weights and enterprise data remain encrypted during processing.  

Other common use cases include secure multi-party computation, where multiple organizations can jointly analyze data without exposing it to each other; privacy-preserving analytics, such as federated learning and collaboration on healthcare data; and financial services workloads like fraud detection. 

They’re also used for cryptographic key management, where keys are generated and used in a tamper-proof environment. Fortanix pioneered the use of Intel TDS enclaves for enterprise-level key management, and that foundation has since expanded to protect full AI workloads on modern GPUs.