Is it now? Adversarial training isn't infinitely scalable either, has its limitations also. Also - the moment that companies start training models to resist d

Is it now? Adversarial training isn't infinitely scalable either, has its limitations also. Also - the moment that companies start training models to resist detectors, they expose themselves to regulation. Won't stop dark AI models running on some website somewhere, but it can be very effectively applied to companies running at Google or OpenAI scale.