The framing here is weird, starting with "Effective Altruism" re-branded as being about nutjobs against AI in the article.

How are AI safety concerns solely about stupid sandboxing issues?