“A human is exposed to approximately 200 million tokens of language by adulthood, whereas frontier AI models are trained on tens to hundreds of trillions of tokens.”