It's like with the Nuclear Bomb, it's not like had Einstein withheld his contributions, we wouldn't have Nuclear Bombs today. It's always a matter of time before someone else figured it out, and until someone else with bad intentions does.
How to attack safety in AI I think has to assume there are already bad actors with super powerful AI around, and what can we do in defense of that.