---
title: "Data Retention"
description: "Content is forwarded at the edge and never stored — a zero-retention architecture designed for GDPR, CCPA and PDPO compliance. No training on your data, ever."
canonical: https://pirouter.ai/data
last-updated: 2026-08-24
---

# Data Retention

> Content is forwarded at the edge and never stored — a zero-retention architecture designed for GDPR, CCPA and PDPO compliance. No training on your data, ever.

PiRouter is an edge LLM gateway: requests enter at the nearest edge node and go straight to the provider — no central region, no disk in the path. What never lands can’t leak, can’t train a model, can’t be sold.

## Lifecycle of a request

1. **Edge ingress** — TLS terminates at the nearest edge node. Authentication and rate limits run right there.
1. **Routing decision** — The best upstream for your model is picked in real time, from live health and latency.
1. **Upstream call** — Content goes straight to the provider — through memory, never to disk.
1. **Response & metering** — The result streams back to you. We keep billing metadata, never content.

## Three commitments

### No content at rest

Prompts and outputs are forwarded, never written to our servers.

### No training

Your data never trains a model. Not ours, not anyone’s.

### No selling

We don’t sell, rent, or share your data. There is nothing stored to sell.

## International data protection

Most platforms bolt compliance on. A zero-retention gateway gets it by construction — the data we process is already the minimum major frameworks demand.

These statements describe how our architecture aligns with each framework’s principles; they are not third-party certifications.

## What we record

Five fields. Billing and reliability, nothing else — never prompts or outputs:

- Field
- Purpose

## Upstream providers

## Questions

---

HTML version: https://pirouter.ai/data · Machine-readable index: https://pirouter.ai/llms.txt