"While unlikely, we cannot rule out that de-identified data derived from their usage of our products helped improve our models." There you go, the suspicion of

"While unlikely, we cannot rule out that de-identified data derived from their usage of our products helped improve our models." There you go, the suspicion of the "concurrent work" (https://cims.nyu.edu/%7Etristanb/statement.pdf https://cims.nyu.edu/%7Etristanb/statement.pdf) mathematicians might not be that unfounded after all...