So the solution is to allow the AI scraping and hide the content, with significantly reduced fidelity and accuracy and not in the original representation, in some language model?