Skip to main content
Silo is a private AI platform. Your conversations are encrypted. We can’t see or store your data. Experience the freedom of truly private, personal AI. Talk naturally, anywhere. Ask Silo for answers, advice, or quick support throughout your day. Private by design. Secure and judgment-free. Ask anything with confidence, knowing your information stays yours forever. Silo is available on iOS, web, and macOS. For users who want no cloud dependency at all, the Silo Box is a hardware server you can run at home.

​
How Silo is different

Most AI apps require sending your conversations to a provider who can log, analyze, or profile your data. Silo brings Signal-level privacy guarantees to personal AI, so you can use the best models without giving up control of your data.
  • Open-source models (DeepSeek R1, Llama 3.3 70B, GLM 4.6/4.7) run inside NVIDIA GPUs with confidential compute mode enabled. This provides hardware-level isolation, encrypted memory, immediate deletion, and cryptographic attestation.
  • Closed-source models are accessed via a privacy proxy. Queries are pooled through proxy servers running in TEEs and stripped of PII by our own Anonymizer model before anything is submitted.
  • Private Deep Research: Multi-step reasoning that runs entirely within a chain of secure enclaves, with only encrypted messages passing between them. The only available private deep research in the market.
  • Private payments via ZCash. Discounted payments via FAI.

​
Models

Completions Model

Use open-source models running in TEEs or closed-source models compatible with the OpenAI API. The completions model is used for chats, memory, and handling the agent.

Embeddings Model

Choose from open-source, closed-source, or local embeddings. Silo comes packaged with a local embeddings model (JinaAI) running on the ONNX inference engine. The embeddings model is responsible for agent memory and semantic search.

Anonymizer Model

The Anonymizer enables you to use advanced closed-source models by performing semantic replacements on requests using a local model trained to understand PII.

STT

Any model compatible with the OpenAI API can be used for speech-to-text. Silo uses Whisper running in a trusted execution environment.

TTS

Any model compatible with the OpenAI API can be used for text-to-speech. Silo uses the Kokoro model running in a trusted execution environment.