See this LessWrong discussion for context: https://www.lesswrong.com/posts/Np5Q3Mhz2AiPtejGN/we-re-not-ready-thoughts-on-pausing-and-responsible-scaling-4?commentId=dz8WsKHGFKN7YxsGX

Anthropic has a number of security commitments in their RSP that they say they need to meet before they develop models they define as "ASL-3 models".

This question will resolve based on whatever the consensus answer between me and Evan (Hubinger) is whether Anthropic did indeed delay scaling or developing more powerful systems for at least one month as a result of those security commitments. This question will resolve whenever Anthropic reaches ASL-4, or when neither me nor Evan think the answer to this is still really going to change (e.g. if Anthropic closes down, or RSPs aren't really a thing anymore, etc.).

If me and Evan disagree I'll ask the following people (in order) to resolve this question one way or another, going down the list whenever someone doesn't respond to me in a day or so, or doesn't want to look into this: Buck Shlegeris, Nate Soares, Holden Karnofsky, Paul Christiano, Mark Xu