Except, they've already done it. They've scraped things they had no permission to scrape. All of their models and weights should be considered public domain at this point.