There are many cases where "anonymous" data can be de-anonymized, mostly if the stats contain outliers or multiple small groups that can be combined to uniquely identify an individual. "Any information that relates to an identified or identifiable living individual" (emphasis mine) implies that if there exists a way to de-anonymize any individual in the dataset then the dataset is PII.
> if there exists a way to de-anonymize any individual in the dataset then the dataset is PII.
Legal department would probably go with,
> if there exists the possibility of a way to de-anonymize any individual in the dataset then the dataset is PII.
Why open themselves to the possibility of lawsuits and fines when they can change the terms.