Pull to refresh
Logo
What Changed
Following Why Sign Up
Judge orders Meta to name employees who approved torrenting pirated books

Judge orders Meta to name employees who approved torrenting pirated books

Rule Changes

Court rules group decisions count: everyone who approved Anna's Archive and LibGen downloads must be named

October 31st, 2026: Former researcher admits torrenting Books3, then deleting copy

Overview

Updated Nov 1

A federal magistrate judge ordered Meta to name every employee who made, authorized, or approved obtaining book collections from shadow libraries for AI training, including who approved using BitTorrent. The October 19 order covers group decisions: if a team agreed to a download, every participant must be identified.

Meta has said it will supplement its answers to address Manohar Paluri, a name the publisher flagged as missing. On October 31, reports said former Meta researcher Stephen Roller admitted in a deposition that he torrented the Books3 dataset while at Meta and later deleted his copy as litigation grew. Meta says the copy sat on his personal computer, not a company server. The June 2025 fair-use ruling left claims that Meta's torrent seeding distributed works to others.

Why it matters

If Meta's torrent seeding is held to be infringement, AI firms that grabbed training data via BitTorrent could face the same liability.

Questions about this story

Free account needed to ask — your question is kept and asked for you right after sign-up. Answers are public.

No questions yet — be the first to ask.

Key Indicators

116.3 TB
Books authors say Meta downloaded
Plaintiffs say Meta downloaded about 116.3 TB between 2022 and 2024: 80.6 TB from LibGen and 35.7 TB from Anna's Archive.
3
Employees named in Meta's first sworn answers
Al-Dahle, Edunov, and Pineau were named for LibGen; Edunov also for Anna's Archive; nobody for Books3.
2022–2024
Years Meta torrented shadow libraries
Entrepreneur Media's brief says Meta 'torrented shadow libraries repeatedly from 2022-2024.'

Voices

Curated perspectives — historical figures and your fellow readers.

Sign up to generate historical perspectives on this story.

Log in Sign Up

People Involved

TH
Thomas S. Hixson
U.S. Magistrate Judge, Northern District of California
Overseeing discovery in the Meta book-copyright cases
 Ahmad Al-Dahle
Ahmad Al-Dahle
Vice president of generative AI, Meta
Named in Meta's sworn answers for the LibGen decision
SE
Sergey Edunov
Director of engineering and research scientist, Meta
Only employee named for the Anna's Archive download
 Joelle Pineau
Joelle Pineau
Vice president of AI research and FAIR head, Meta
Named in Meta's sworn answers for the LibGen decision; Meta has said it will supplement its answers to address Manohar Paluri, who was not named
XW
Xiaolan Wang
Engineer, Meta
Ran the April 2024 Anna's Archive download
NB
Nikolay Bashlykov
Former engineer, Meta
Carried out torrent downloads of LibGen content
 Richard Kadrey
Richard Kadrey
Lead plaintiff, Kadrey v. Meta
Class action narrowed to BitTorrent claims
SR
Stephen Roller
Former AI researcher, Meta (FAIR)
Testified he torrented Books3 during his Meta tenure and deleted his copy

Organizations Involved

Meta Platforms
Meta Platforms
Technology company
Defendant in book-copyright cases

Built Llama 1 and 2, the models at the center of the training-data lawsuits.
Anna's Archive
Anna's Archive
Shadow library
Source of training data at issue

A pirate library cataloging books and scholarly papers, torrented by Meta.
LG
Library Genesis (LibGen)
Shadow library
Source of training data at issue

A shadow library of books and scholarly papers, including the Scimag collection.
Entrepreneur Media
Entrepreneur Media
Publisher
Plaintiff in a separate contributory infringement suit

Publishes Entrepreneur magazine; accuses Meta of seeding its works via BitTorrent.
U.S. District Court, Northern District of California
U.S. District Court, Northern District of California
Federal court
Hearing the Meta book-copyright cases

The federal court in San Francisco and Oakland where the suits proceed.

Timeline

2023 October 2026

6 events Latest: October 31st, 2026 · 1 week ago
Tap a bar to jump to that date

  1. Former researcher admits torrenting Books3, then deleting copy

    Latest Legal

    Reports said Stephen Roller, a former Meta AI researcher, testified he downloaded Books3 by torrent during his 2017–2023 tenure and deleted his copy as litigation mounted. Meta says the copy was on his personal computer and does not show the download served Meta.

  2. Judge orders Meta to name all torrenting decision-makers

    Ruling

    Hixson rules every team member in a group approval must be identified, including seeding decisions. The order also sets a seven-day deadline for internal training-data catalog documents.

  3. Meta identifies three employees in sworn answers

    Legal

    Meta names Al-Dahle, Edunov, and Pineau for LibGen; Edunov for Anna's Archive; nobody for Books3.

  4. Meta ordered to produce torrent server logs

    Legal

    Hixson ordered Meta to hand over command-history files from every server it used to torrent.

  5. Fair-use ruling narrows the case

    Ruling

    Judge Vince Chhabria rules Meta's training on the authors' books was fair use on the arguments before him, leaving torrenting claims live.

  6. Authors sue Meta over book training data

    Legal

    Kadrey, Silverman, and others file suit over Meta's use of copyrighted books in Llama training.

Scenarios

1

Meta names more employees — or swears it can't

Likely Resolves by Nov 30, 2026

Discussed by: Authors' counsel and Entrepreneur Media, who sought the order

Meta must answer interrogatory 22 naming every person who made, authorized, or approved the torrenting decisions, including everyone in group decisions. If senior figures beyond the three already named surface, authors get sworn testimony tying specific executives to the downloads. If Meta instead swears no one else decided, plaintiffs can depose the named executives anyway.

2

Meta challenges the discovery order

Possible Resolves by End of 2026

Discussed by: Meta's counsel, which has resisted naming executives

Meta argued decisions were informal and collaborative, and said it knew of no one who approved uploads or distribution. It could object to Hixson's ruling, seek a protective order, or ask the district judge to review under the Federal Rules. A challenge would delay answers and could narrow the order's scope.

3

BitTorrent seeding claims head to trial or settle

Uncertain Resolves by End of 2027

Discussed by: Court filings in the Kadrey and Entrepreneur Media cases

The fair-use ruling left only the distribution claims alive. Server logs and named decision-makers could let authors show Meta seeded works to other torrent users, supporting contributory infringement. A summary judgment ruling, settlement, or trial on those claims would decide whether torrenting training data carries copyright liability.

4

Roller testimony forces Meta to revise its Books3 answers

Possible Resolves by End of 2026

Discussed by: Plaintiffs, who say the testimony contradicts Meta's August statement that it did not obtain Books3 through BitTorrent

Roller's deposition may prompt plaintiffs to seek supplemental sworn answers on who obtained Books3. Meta has disputed that the testimony shows the download served Meta.

Historical Context

3 moments from history that rhyme with this story — and how they unfolded.

2001

A&M Records v. Napster (2001)

Napster let millions share MP3 files peer-to-peer, both uploading and downloading copyrighted songs. The Ninth Circuit held Napster liable for contributory and vicarious infringement, rejecting the fair-use defense Napster raised for its users.

Then

Napster shut down and settled with record labels, paying millions in damages.

Now

The ruling established that platforms enabling file sharing can be liable even when they don't host files, drawing a line between private copying and public distribution.

Why this matters now

Meta's torrenting involves the same uploading (seeding) versus downloading distinction the Napster court weighed, and contributory infringement is the claim authors now press against Meta.

October 2015

Authors Guild v. Google (2015)

Google scanned millions of library books without permission to build its search index and display snippets. The Second Circuit ruled the scanning was transformative fair use, rejecting the authors' case.

Then

Google kept its book search; the authors' damages claims failed.

Now

The case became the anchor precedent that large-scale copying of books can be fair use when the purpose is transformative.

Why this matters now

Judge Chhabria's June 2025 ruling that Meta's training on books was fair use drew on this tradition, which is why only the BitTorrent distribution claims remain.

March 2023

Hachette v. Internet Archive (2023)

The Internet Archive scanned physical books and lent them digitally through controlled digital lending, arguing it was fair use. A district court ruled against it, finding the lending harmed publishers' ebook markets; the Second Circuit affirmed in 2024.

Then

The Internet Archive removed the contested books from lending.

Now

The ruling narrowed where fair use protects mass copying of entire works.

Why this matters now

It shows courts scrutinize the purpose of mass copying, and the seeding claims ask whether Meta's distribution to other torrent users crosses the fair-use line that protected its training.

Sources

(13)
+7
Slashdot Ars Technica Digital Music News The Hum India Today GIGAZINE TokenPost Nydus U.S. Government Publishing Office Infobae Mad Robot LavX News BetaNews