SAN FRANCISCO (LN) — U.S. Magistrate Judge Susan Van Keulen on Thursday modified a court-established discovery protocol in Doe I et al v. Google LLC, capping the number of data sources plaintiffs can demand at 1,424. The order reduces the plaintiffs' initial selection of 7,265 sources by approximately 80%.

The dispute centers on Step 2 of a protocol designed to facilitate the production of data source information in the litigation. Under the protocol, Google must list data sources, plaintiffs must identify a relevant subset, and Google must produce field-level data for those sources.

At Step 2, plaintiffs selected approximately 7,265 data sources from Google's initial list of 18,088 sources across three databases: Kansas/Oz, Sawmill, and Spanner/Big Table/Napa.

Google argued that producing field names and descriptions for the plaintiffs' selected subset would be disproportional, noting that some sources "may have hundreds of thousands of fields each," potentially leading to the production of millions of fields.

"The universe of potentially relevant data is vast, as is the universe of potentially irrelevant data," Van Keulen wrote. "This is the very issue addressed by the Protocol."

The judge modified the protocol to cap the total number of data sources at 200 for the Sawmill and Spanner/Big Table/Napa databases combined, down from the plaintiffs' selection of more than 7,000 for those specific databases.

Under the new order, plaintiffs must identify seven "key" Sawmill logs for full field descriptions and comments, plus up to 700 additional Sawmill sources for field names only. They may also select up to 700 data sources from the Spanner/Big Table/Napa database.

The amended total of 1,424 data sources represents a significant reduction from the plaintiffs' original request.

Plaintiffs have until May 24, 2026, to submit their updated selections. The parties may agree to a modest extension of this deadline.

Van Keulen declined to set a cap for Step 4 of the protocol at this time, relying on the parties to proceed under the proportionality requirements of Federal Rule of Civil Procedure 26(b)(1).

The case is Doe I et al v. Google LLC, Case No. 3:23-cv-02431-VC (SVK).