AI Safety Action Plan - A report commissioned by the US State Department

This morning, Gladstone AI published a report they made that was commissioned by the US State Department. The report not only acknowledges catastrophic risks from AI, but also includes a very ambitious set of proposals designed for the US to mitigate catastrophic risks arising from advanced AI, going from export controls to a new federal regulatory agency, and even suggesting the creation of multilateral treaties for a new international regulatory agency.

I've only read the executive summary and skimmed the full report, but it looks pretty great. It's detailed and concrete, it's written very well, and includes some pretty useful diagrams to explain their implementation proposals. It's surprisingly polished for a report made by just a few authors.

The report also got pretty extensive coverage by Time Magazine today (which published a general article, as well as one zooming in on the expert interviews). You can also find a good Twitter thread of the lead author explaining the context for the report here.