No. The point is that the attack surface is more vast for data-driven models.