Comment by the_mar

but the argument here is that the ai company does not explicitly train agents to hack, it’s that it doesn’t prevent them from doing so sufficiently. so it’s negligence and not malicious intent