Jonathan Mortensen is the founder and chief executive of Confident Security, which develops verifiably private AI inference for organizations handling sensitive data. His approach replaces reliance on providers’ retention policies with infrastructure that limits access to prompts and makes its security guarantees independently inspectable.
Mortensen studied biomedical engineering at Case Western Reserve University and earned a doctorate in biomedical informatics from Stanford in 2016. After working in healthcare data science at Nuna, he co-founded Gyroscope Software, which was acquired by BlueVoyant, and led West Coast data science at the cybersecurity company.
He subsequently founded bit.io, a serverless PostgreSQL platform, and became a staff software engineer at Databricks after its acquisition. There, his work encompassed multi-tenant security, encryption, infrastructure isolation, and data ownership. As a South Park Commons Founder Fellow, he turned to protecting confidential information processed by remote AI systems.
Founded in 2024, Confident Security emerged publicly in July 2025 with $4.2 million in seed funding. Its commercial service, CONFSEC, complements OpenPCC, an open-source framework for adapting private cloud inference beyond proprietary consumer ecosystems.
- Privacy enforced by architecture. Stateless processing, encrypted requests, and restricted administrative access constrain what an operator can retain or inspect; contractual promises alone cannot provide equivalent technical assurances.
- Remote attestation and software transparency. Clients can inspect cryptographically authenticated claims about a server’s hardware and software before disclosing sensitive information. Append-only transparency logs enable independent scrutiny of deployed code and help expose unexpected software changes.
- Protection for identity and metadata. Encrypted prompts still leave users vulnerable if authentication and routing identify them. Oblivious HTTP, blind-signature authentication, and unlinkable requests separate request contents from user identity and make individual targeting harder.
- Private cloud computing outside Apple’s ecosystem. Mortensen’s independent analysis of Apple’s Private Cloud Compute informs OpenPCC’s combination of open-source software, trusted hardware, confidential computing, and reproducible builds. He also emphasizes the operational costs: additional latency, encryption overhead, complicated clients, limited debugging, and the difficulty of measuring usage without identifying users.
