The Unofficial Guide to Apple’s Private Cloud Compute
AI Engineer World's Fair 2025 · 20:36
Private AI inference infrastructure
Confident Security builds infrastructure for companies that want to use AI models with confidential data. Its CONFSEC managed service is based on OpenPCC, an open-source framework for running open or custom models. The framework is designed to protect prompts, outputs and logs through encrypted streaming, hardware attestation and unlinkable requests, making privacy protections available for inspection and audit.
Founder Jonathan Mortensen previously founded and sold Gyroscope to BlueVoyant and bit.io to Databricks. Confident Security’s technical approach draws inspiration from Apple’s Private Cloud Compute. OpenPCC uses Oblivious HTTP with a third-party relay to prevent compute providers from learning about individual user actions. In 2025, the company joined the Confidential Computing Consortium, contributing to an ecosystem focused on protecting data during computation.
Developers can work with the managed service or deploy the OpenPCC framework on their own infrastructure. CONFSEC’s documented onboarding includes a web console, service accounts with optional spending limits, API keys with expiration dates and a downloadable command-line client. OpenPCC also provides a Go client and a C library underlying Python and JavaScript clients.
AI Engineer World's Fair 2025 · 20:36
Affiliations reflect their AIE appearances, not necessarily current employment.
Mortensen explains how remote attestation and a public software transparency log establish verifiable trust, alongside requirements for stateless processing and non-targetability.
Affiliations reflect each recorded session, not necessarily current employment. Mortensen’s presentation is an independent guide; he states that he does not work for or represent Apple.