In both cases a central safety effort seems nearly impossible. E.g. trying to enforce international AI risk cooperation via air strikes against data centers [1]

In both cases a central safety effort seems nearly impossible. E.g. trying to enforce international AI risk cooperation via air strikes against data centers [1] can easily be avoided by defecting countries by building supercomputers underground. With the moon bugs this wasn't a big problem, as they were so unlikely. But for AI the risk seems quite large to me. [1] https://time.com/6266923/ai-eliezer-yudkowsky-open-letter-not-enough/ https://time.com/6266923/ai-eliezer-yudkowsky-open-letter-no...