> While unlikely, we cannot rule out that de-identified data derived from their usage of our products helped improve our models This is covering for Tristan sa

> While unlikely, we cannot rule out that de-identified data derived from their usage of our products helped improve our models This is covering for Tristan saying something like, "Actually, I was using my friend's account for half of this work".