Archive - Redwood Research blog

July 2025

Should we update against seeing relatively fast AI progress in 2025 and 2026?

Maybe we should (re)assess the case for relatively fast progress after the GPT-5 release.

Jul 28 •

Ryan Greenblatt

Why it's hard to make settings for high-stakes control research

It's like making challenging evals, but more constrained

Jul 18 •

Recent Redwood Research project proposals

Empirical AI security/safety projects across a variety of areas

Jul 14 •

Ryan Greenblatt

,

,

,

,

, and

What's worse, spies or schemers?

And what if you have both at once?

Jul 9 •

and

Ryan on the 80,000 Hours podcast

Ryan’s podcast with Rob Wiblin has just come out!

Jul 8 •

How much novel security-critical infrastructure do you need during the singularity?

And what does this mean for AI control?

Jul 5 •

Two proposed projects on abstract analogies for scheming

We should study methods to train away deeply ingrained behaviors in LLMs that are structurally similar to scheming.

Jul 4 •

There are two fundamentally different constraints on schemers

"They need to act aligned" often isn't precise enough

Jul 2 •

June 2025

Jankily controlling superintelligence

How much time can control buy us during the intelligence explosion?

Jun 27 •

Ryan Greenblatt

#nojs-banner { position: fixed; bottom: 0; left: 0; padding: 16px 16px 16px 32px; width: 100%; box-sizing: border-box; background: red; color: white; font-family: -apple-system, "Segoe UI", Roboto, Helvetica, Arial, sans-serif, "Apple Color Emoji", "Segoe UI Emoji", "Segoe UI Symbol"; font-size: 13px; line-height: 13px; } #nojs-banner a { color: inherit; text-decoration: underline; } This site requires JavaScript to run correctly. Please turn on JavaScript or unblock scripts