Putting GenAI on Notice: GenAI Exceptionalism and Contract Law

Gathering enough data to create sufficiently useful training datasets for generative artificial intelligence requires scraping most public websites. The scraping is conducted using pieces of code (scraping bots) that make copies of website pages. Today, there are only a few ways for website owners to effectively block these bots from scraping content. One method, prohibiting scraping in the website terms of service, is loosely enforced because it is not always clear when the terms are enforceable. This paper aims to clear up the confusion by describing what scraping is, how entities do it, what makes website terms of service enforceable, and what claims of damages website owners may make as a result of being scraped. The novel argument of the paper is that when (i) a site's terms of service or terms of use prohibit scraping or using site content to train AI and (ii) a bot scrapes pages on the website including those terms, the bot's deployer has actual notice of the terms and those terms are therefore legally enforceable, meaning the site can claim a breach of contract. This paper also details the legal and substantive arguments favoring this position while cautioning that nonprofits with a primarily scientific research focus should be exempt from such strict enforcement.