They didn’t deny it, they said if the other workers’ research ended up in the training data it wasn’t deliberate. Then they gave the researcher who noticed the similarities a “co-publish with us or else” offer.