I feel like this is so they can deny that they fed all the webpages that they cached to their ‘AI’ training datasets later when someone accuses them of that. Now when asked about the copies of webpages that they have they can be like “What copies?” and end the conversation there.