"The last thing we need is to have algorithms designed to maximize revenue driving LLM output and interactions," Cuban wrote on X.