I've contributed to AI Safety field-building in Germany over the past few years (ML4Good bootcamps, AI Safety Collab). The idea of centralized German AIS infrastructure has been discussed many times over the years - SAIGE is the most well-prepared and sensible attempt I've seen so far.

I've reviewed their planning documents and I'm impressed: thorough failure modes analysis, smart partnerships with existing orgs rather than reinventing the wheel, strong team with good connections to the international ecosystem. Jessica seems like a good leader for this.

No guarantees of success, but definitely worth a try - high expected value :)

Atlas is great. I think there should be additional courses with a different style then the BlueDot courses that are a bit lengthy etc. I think Atlas could produce that!