The AI API with zero
data exposure
With Privatemode, you get on-prem like privacy combined
with the convenience and flexibility of the cloud.
Trusted by security‑critical teams at enterprises and public institutions
Product
Overview
Use cloud‑grade AI with
on‑prem‑level confidentiality.
Models
Use frontier‑grade LLMs
without giving up privacy.
Privatemode gives you access to leading open-weight models through one secure endpoint, so you can choose the best model for each workload without exposing sensitive data.
Performance
Get cloud‑level speed on
sensitive workloads.
The inference API runs on modern GPU infrastructure to deliver low‑latency responses and high token throughput, even under heavy load. You keep real‑time performance for agents and user‑facing apps without managing your own GPU clusters.
Security and encryption
Privatemode protects
your data end-to-end.
Privatemode layers end‑to‑end encryption with confidential computing and verifiable attestation, so you can verify where and how your data is processed. Explore the security architecture to see how key management and attestation defend against both external attacks and insider access.
Integration
Keep your OpenAI and Anthropic integrations,
change only the endpoint.
The Privatemode Encryption Proxy is fully compatible with the OpenAI and Anthropic API, including existing SDKs and client libraries. Point your apps to the proxy URL and reuse your code while all traffic is transparently encrypted end‑to‑end.
Pricing
Pay only for tokens you
actually use.
Predictable usage-based pricing with no hidden fees, from free tier to production. Enterprise plans with custom rate limits available on request.
Introduction
Confidential inference API
Use cloud‑based AI without losing control of your data.
Drop‑in, OpenAI‑compatible API
Run the Privatemode Encryption Proxy and keep your existing OpenAI clients, SDKs, and request formats—just point them to the new endpoint. Your apps talk to a familiar API while the proxy transparently encrypts prompts and decrypts responses for you.
Your data never leaves your control
Prompts and responses are encrypted before they leave your infrastructure and stay encrypted in transit and at rest. Only the AI inside Privatemode’s confidential‑computing environment can access plaintext, never operators or cloud providers.
Confidential computing, end‑to‑end
Privatemode uses hardware‑based confidential computing to keep data encrypted even while it is processed in main memory. Remote attestation verifies the runtime and models before any request is decrypted, so only approved code ever sees plaintext.
How to get started
In 3 simple steps
As easy as 1-2-3.
As secure as it gets.
docker run -p 8080:8080 ghcr.io/edgelesssys/privatemode/privatemode-proxy:latest --apiKey <your-api-key>Run the Privatemode Encryption Proxy.
The proxy verifies the integrity of the Privatemode service using confindential computing-based remote attestation. The proxy also encrypts all data before sending and decrypts data it receives.
#!/usr/bin/env bash
curl localhost:8080/v1/chat/completions \
-H "Content-Type: application/json" \
-d '{
"model": "gpt-oss-120b",
"messages": [
{
"role": "user",
"content": "Hello Privatemode!"
}
]
}'Send your prompts to the proxy.
You now have programmatic access to an end-to-end secure AI. The proxy is compatible with the OpenAI Chat API. Process your own sensitive data or provide trustworthy services to others – the possibilities are endless.
Start using the Privatemode API
Sign up for free and get your API key right away. Works in most tools that work with OpenAI or Anthropic.
Comparison
The scalability of the cloud.
The security of on-prem.
- Data Privacy
- Compliance
- Setup time
- Costs
- Maintenance
- Scalability
Privatemode
Cloud-grade AI with end-to-end encryption and confidential computing. Built for industries where compliance isn't optional.
- Confidential computing
- Compliance by design
- Ready within minutes
- No infrastructure costs
- Fully managed
- Automatic
ChatGPT
Powerful AI, but prompts and data are processed unencrypted on OpenAI's servers, with no verifiable privacy guarantees
- Contractual
- Limited
- Usable within minutes
- No infrastructure costs
- Fully managed
- Automatic
Self-hosted AI
Full data control, but requires dedicated hardware, in-house ML ops, and ongoing maintenance.
- Full control
- Full control
- Weeks to months
- High upfront
- Self-managed
- Manual
Case Studies
Explore how teams integrate Privatemode into their workflows
Case Study
The German Federal Employment Agency
The German Federal Employment Agency deploys cloud-based AI with Privatemode.
Case Study
Kauz.ai
Kauz.ai leverages Privatemode to provide confidential chatbots to customers in regulated industries.
Case Study
HÄVG
Germany's General Practitioners' Contract Association piloted Privatemode's confidential computing platform to build a privacy-compliant AI chat.
FAQ
Technical Details
Frequently asked questions about using Privatemode API
Install Docker, run the Privatemode proxy with your API key, and start sending requests to localhost:8080. The proxy handles all encryption and attestation automatically. Full instructions are in the quickstart guide — setup takes around 10 minutes.
Want to see Privatemode in action?
We're happy to show you around and give an overview of what's possible.