Redwood Research

Redwood Research is a small non-profit that works on AI alignment and, in particular, on what the field calls control: the question of how to keep a powerful model from causing harm even if it is actively trying to, rather than assuming you can make it want the right things in the first place.

It shows up in news coverage mostly as an outside reviewer. Along with METR, it was brought in to independently investigate the misalignment behind OpenAI’s July 2026 Hugging Face incident. That role, an outside group checking a lab’s homework, is one of the few forms of accountability the industry currently has.