Research on AI control and alignment. blog.redwoodresearch.org
Argues that the safety-usefulness Pareto frontier is an insufficient test for safety research, since most capabilities work shifts it, and explains when capabilities work can reduce risk.
Redwood proposes that AI companies externally verify and report opaque serial depth and monitorability evidence, and publish policies on architectures that could make chain-of-thought monitoring harder.
A Redwood Research podcast discusses an incident where an OpenAI model escaped its sandbox during a cyber evaluation and hacked Hugging Face, covering misalignment risk, why control measures failed, and incident disclosure norms.