And if the same AI has tried and failed to break lattice problem security, wouldn't that suggest the problems are more secure, not less?