Tinfoil

About Tinfoil

Tinfoil builds verifiably private AI — the framing on the site is that you should be able to trust the hardware rather than a vendor's pinky-promise. Workloads run inside secure enclaves so that neither Tinfoil nor the underlying cloud can read the data being processed, and the setup is designed to be independently verifiable rather than merely asserted in a privacy policy. Three products sit on that foundation. Private Chat is a consumer-facing assistant available in the browser and on iOS for $20 a month, including up to 3M tokens per hour under a fair-use policy, web search, projects, document upload, speech to text, and sync across devices. Private Inference is an OpenAI-compatible, open-source, fully verifiable API priced by usage, so an existing application can point at it with a base-URL change. Tinfoil Containers lets you run any Docker image inside an enclave at $20 a month plus usage, which is the piece that lets teams bring custom AI workloads rather than only calling hosted models. The company is SOC 2 compliant and open source, has been featured in a Llama case study as the only multi-GPU offering with production-ready verifiable privacy, and collaborates with Red Hat and Ubuntu on confidential-computing infrastructure.

Key Features

Tinfoil Pros & Cons

Tags