How Ubuntu enables verifiably private AI with confidential computing on bare metal
Tinfoil’s success story
Download the full guide ›
As a US-based cybersecurity company focused on verifiable private AI, Tinfoil needed to guarantee that sensitive data remains encrypted and private at every stage of processing, even from the platform provider itself. Their platform ensures financial records, health information, legal documents, and proprietary models are processed exclusively inside hardware trusted execution environments (TEEs). But supporting heterogeneous infrastructure across AMD, Intel, and NVIDIA hardware without custom kernels or cloud lock-in required a fully open, audit-ready operating system with native confidential computing support.
Tinfoil selected Ubuntu 25.04 for its official AMD SEV-SNP and Intel TDX enablement, reproducible build tooling, and proven security pedigree. Ubuntu was the first production-grade OS to ship native support for both AMD SEV-SNP and Intel TDX without patched kernels or experimental branches. With Ubuntu’s compatibility with mkosi and automated CI, Tinfoil built a single, attested image deployable across both AMD and Intel hardware. Every image is constructed in CI, cryptographically measured, and published publicly for independent client-side verification before any sensitive data is processed.
“Ubuntu provides the strongest foundation for building trustworthy, multi-platform confidential-computing workloads, allowing Tinfoil to build a verifiably private AI platform with minimal custom code, ensuring maximum auditability.”
— Tanya Verma, Co-Founder, Tinfoil
Read the guide to learn:
- Why Tinfoil chose Ubuntu 25.04’s native AMD SEV-SNP and Intel TDX support for verifiable private AI
- How Ubuntu’s audit-ready build infrastructure and reproducible snapshots enabled cryptographically verifiable supply chain transparency
- How enterprises in finance, healthcare, and legal sectors can adopt AI without transferring trust to the AI operator or cloud provider
Further reading