← home

PRIVACY & AI

Why private AI matters

And what "private" actually means when it comes to your AI agent.

The problem with shared AI platforms

When you use a shared AI platform, your conversations, documents, and data pass through infrastructure you don't control. In most cases:

For personal use, this might be acceptable. For business use — handling contracts, financial data, health records, legal documents, or proprietary IP — it's a liability.

What "private AI" actually means

Private AI isn't just a marketing label. It means specific, verifiable technical properties:

1. Dedicated infrastructure

Your agent runs on a VPS that belongs to you alone. Not a shared container. Not a "namespace" on someone else's cluster. A dedicated server with its own CPU, RAM, and disk — isolated at the hardware level.

2. Full-disk encryption with your key

LUKS full-disk encryption means all agent data — conversations, memory, files — is encrypted at rest. With Bring Your Own Key (BYOK), you hold the only decryption key. If we got subpoenaed, we'd hand over a brick.

3. Zero-trust networking

Zero open ports. No public attack surface. Access only via Tailscale VPN — a private mesh network where every connection is authenticated. Your server is invisible to anyone not on your network.

4. No data exfiltration path

We don't have access to your server. We don't have your encryption key. We don't log your conversations. There is no path — technical, legal, or social — for your data to leave your control.

The subpoena test
Ask your current AI provider: "If you received a subpoena for my data, what would you hand over?" If the answer is anything other than "encrypted data we can't read," your data isn't private.

Who needs private AI?

Lawyers and legal teams. Attorney-client privilege doesn't apply if a third party has access to the conversation. Using a shared AI platform to draft legal documents means that privilege may be waived.

Healthcare organizations. HIPAA requires that protected health information (PHI) is stored with appropriate safeguards. A shared AI platform where staff can review conversations doesn't meet that bar.

Financial services. Client financial data, trading strategies, and deal flow information are tightly regulated. Shared infrastructure is a compliance risk.

Startups and R&D teams. Proprietary code, product roadmaps, and competitive intelligence shouldn't be training data for someone else's model.

Anyone who values autonomy. You wouldn't store confidential documents on a stranger's laptop. Why store your AI conversations on a stranger's server?

The EU advantage

All RunYourAgent servers run in Hetzner's EU data centers (Germany and Finland). This means:

This matters especially for European businesses that need to comply with GDPR, but also for anyone who wants the strongest possible legal framework protecting their data.

How RunYourAgent works

We provision a dedicated, encrypted VPS for each customer. The stack:

Provisioning takes about 5 minutes. You pick your agent framework, choose your security tier, and deploy. Once running, you connect via SSH over Tailscale or chat with your agent on Telegram.

"The best privacy guarantee is a technical one: we physically cannot access your data, even if we wanted to."

What we don't do

Getting started

RunYourAgent starts at EUR 9.99/month for a Starter instance (3 vCPU, 4 GB RAM, 80 GB encrypted SSD). All tiers include zero-trust networking, LUKS encryption, and CIS hardening.

deploy your private agent →