
It’s basically an admission that their guardrails are more like a stack of Swiss cheese than anything deterministic to stop requests of those kind. Each layer of the cheese has holes, but stacked one on top of the other, catch most things. However there can still be holes through all the guardrails that they haven’t yet covered up, and $500k per incident could be potentially devastating financially. Probably every lab is thinking this btw, Elon is just the only one who cares so little about PR