New Collaboration: Shallow Review of Technical AI Safety, 2025

We recently collaborated with the Arb Research team on their latest technical AI safety review. This document provides a strong overview of the space, and we built a website to make it significantly more manageable.

The interactive website: shallowreview.ai

The review examines major research directions in technical AI safety as of early 2025, including mechanistic interpretability, scalable oversight, and various alignment approaches. It's designed as an accessible entry point for researchers wanting to understand the current landscape.

One particular challenge for the website was differentiating the numerous sections. The field has substantial parallel work happening across many domains, making it difficult to maintain orientation. We color-coded each section for distinctiveness and removed visual clutter to keep the core content in focus.

The site also features a table view and a customizable cluster diagram.

This represents our continued work supporting epistemics infrastructure around the AI safety community. We hope people find it useful!