Welcome to simpleaudit Discussions!

[kelkalot]

👋 Welcome to the SimpleAudit community!

SimpleAudit is an open-source, local-first framework for red-teaming and safety-auditing AI systems. It's developed by Simula and SimulaMet with the Norwegian Directorate of Health, and it's a verified Digital Public Good. This is the place to ask questions, share what you're auditing, and help decide where the project goes next.

Where to post

• 🙏 Q&A: setup, providers, judges, reading scores. Please mark the answer that solved it.
• 💡 Ideas: new features, judges, integrations.
• 🧪 Scenario packs: propose or share scenarios for new languages, domains or sectors. Many of the built-in packs came from real public-sector audits, and yours could become the next one.
• 🙌 Show and tell: audits you've run, model comparisons, and blog posts or papers that use SimpleAudit.
• 📣 Announcements: releases and news from the maintainers.

🐛 Bugs go in Issues so we can track the fix. 🔒 Please report security problems as described in SECURITY.md, not in public posts.

Introduce yourself 👇

Reply below and tell us:

• who you are, and where you work if you'd like to share
• what you're auditing (a model, chatbot, RAG app or agent) and in which language
• one thing you wish SimpleAudit did

Ground rules

Be kind and assume good faith (see our Code of Conduct). Before you share a transcript, remove any harmful model output, personal data or API keys.

Useful links

• Quickstart
• Guidelines for writing scenarios
• Paper: arXiv:2605.06652
• Live demo

If SimpleAudit is useful to you, a ⭐ on the repo helps others find it.

The SimpleAudit team (Michael, Sushant and contributors)