A deepfake bot has been generating explicit, non-consensual images on the platform. The researchers who found it say their warnings have been ignored.