> While unlikely, we cannot rule out that de-identified data derived from their usage of our products helped improve our models . Wow. That sounds like an admi

> While unlikely, we cannot rule out that de-identified data derived from their usage of our products helped improve our models . Wow. That sounds like an admission of guilt.