About HatCat

HatCat is an open-source interpretability framework that enables real-time concept detection and steering in language models. It's part of a larger vision—the Fractal Transparency Web—for building AI governance infrastructure that doesn't depend on any single point of control.

The code is CC0 (public domain). You're not just allowed to fork it and build your own versions—we're counting on you to. The defense thesis depends on diverse lenses from diverse perspectives.

Created by Possum Hodgkin - Experience Architect for AI Governance

EU and African AI Governance Kaouthar El Bairi - AI Safety Researcher

With thanks to:

• MIT Futures Team — for advanced access to AI Risk Repository 2026 report
• CSIRO Software Systems Research Group — for access to their AI safety researchers
• Swiss AI — for releasing enough detail in Apertus to make this possible
• The R&D teams from my.gov.au and servicesaustralia.gov.au — for sense checking and testing
• GovCMS and GovAI — for providing so much support and compute
• Claude and Anthropic — for contributing so much, and making that possible