Earlier quoted context omitted.
Anonymized data generally is not very anonymous. You can only trust that data is actually anonymous if there exists no non-anonymous dataset with which you can cross reference the anonymized data, and you can guarantee that no such database will ever exist. I'm willing to bet that many "anonymized" datasets are of the kind which could relatively easily be deanonymized, especially datasets which are frivolously used t…
That's an incredibly broad statement; we are talking about what movies people have watched on netflix. What possible non-anonymous dataset could you be cross referencing to de-anonymize it? Even if you could, who cares?
> what is harmful about using anonymised data this way?
Whether it's harmful to use such data when it's not anonymous is a different discussion.