# ISNAD: An open-source, self-hosted tool for tamper-evident AI audit logging (EU AI Act Art 12 ready) — retained record

> Retained reddit record from reddit:selfhosted.

Canonical: https://highsignal.app/data/records/fd68295a929ef09f

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:selfhosted)  published 2026-10-08 18:41 ingested 2026-10-09 02:37

[original source ↗](https://www.reddit.com/r/selfhosted/comments/1x0z15r/isnad_an_opensource_selfhosted_tool_for/)

## retained text



```
Hey everyone,

I've been working on a tool called ISNAD for the self-hosted community. It's an open-source (Apache-2.0) engine that creates a locked, cryptographically signed record of an AI's answer—showing who or what handled it at each step.

The goal is to provide "proof of custody" for AI decisions. Instead of just a standard log, you get a tamper-evident record that you can verify offline. This is especially useful for anyone dealing with the EU AI Act's upcoming logging requirements (Art 12 & 19).

It runs entirely on your own infrastructure, so your data never leaves your control.

You can `pip install isnad` and find the repo here: [https://github.com/alizahidraja/isnad](https://github.com/alizahidraja/isnad)

Curious to hear what the self-hosted community thinks about the architecture and approach.
```



document fetched 2026-10-09 00:31 · canonical https://reddit.com/r/selfhosted/comments/1x0z15r/isnad_an_opensource_selfhosted_tool_for
