1. Privacy/data ownership. So I realize data is going to be anonymized, but still, Zotero's privacy policy, which to me is one of its core strengths, is to "not share your data with third parties."

We've given this a lot of thought over the years, and ultimately we don't consider sharing these sorts of aggregate statistics to constitute sharing private data. We don't feel there's an articulable privacy concern with these counts — and certainly not one that justifies the vastly less useful data that would result from limiting the counts to public libraries, which represent a tiny fraction of total Zotero libraries.

The page you quote also discusses the use of "anonymous and aggregated" data for "research and analysis", which, though it doesn't explicitly mention sharing, suggests this sort of usage.

While we therefore see this usage as covered by our existing policies, we're going to be rolling out updated terms of service and an updated privacy policy this week for a variety of reasons (mainly to consolidate all Zotero software and services into a single document), and the updated privacy policy makes this new usage more explicit. (We'll be notifying people of the changes via email, since our current ToS say we will.)

To be clear, though, this aggregate data isn't yet available, and if anyone is concerned with this usage, they can of course delete their data from the Zotero servers.

2. To what extent do you think the new API will be useful in handling retrieve metadata requests?

It's separate, at least at this point. Currently for this grant we're only talking about providing readership counts for given identifiers (DOI, ISBN, etc.). A lookup service for metadata retrieval would be a separate API that took full-text queries and returned identifiers.

Item counts per identifier do indeed seem entirely unproblematic.

OK, I'm glad you agree on this — and that is indeed all we're talking about here.

We would not expose metadata in private libraries, anonymized or not. If, for example, metadata were provided along with these readership counts, it would come solely from public sources, even if the aggregate readership counts themselves included items in private libraries.

No, I was referring to the metadata of future "global items" made available from the Zotero API and zotero.org — that is, metadata accompanying these readership counts. I'm saying that, while the readership counts may include items in private libraries, any metadata displayed would come from either external sources, resolved using those items' UUIDs (ISBNs, DOIs, etc.), or, for items without UUIDs, matched items in public Zotero libraries. So metadata on private items will never be exposed.

The datasets described in the grant announcement will follow these same restrictions — we're not providing anything that won't eventually be available publicly via the Zotero API. (Part of the purpose of the grant is just to improve the accuracy of that data before it's released.) We'll essentially be providing identifier:count:public-user-items, where 'identifier' is a UUID (or, where there isn't one, a Zotero global item id), 'count' is a readership count (across all Zotero libraries), and 'public-user-items' is the set of constituent user/group items in public libraries. No data on private items will be included beyond the readership counts.

We're aware that "anonymized" is a bit of a scary word, particularly in tech circles. In this context we really just mean "aggregate", and anonymous simply to the extent that readership counts would be inherently anonymous.

they're working on it but nothing that's been made public, no.

Any news on this? Like wouterhendrickx above, I would love to see Zotero represented in altmetrics. (I've seen this: https://www.zotero.org/blog/studying-the-altmetrics-of-zotero-data/ — but haven't seen anything since either here or on the altmetrics/impactstory side.)

which recently has made a commendable turn towards fully open data.

and is now free again for researchers to use.