According to TechCrunch, a federal judge gave final approval on July 20, 2026 to Anthropic's $1.5 billion class-action copyright settlement, described as the largest in U.S. copyright law history, covering an estimated 500,000 works at roughly $3,000 each.
When Was the Settlement Approved, Who Signed It, and What Is Its Historic Standing?
According to TechCrunch, which cited Reuters, a federal judge gave final approval on Monday, July 20, 2026, to Anthropic's landmark $1.5 billion settlement of a class-action copyright lawsuit. The order was not signed by the judge who originally handled the case, William Alsup, because he has since retired. Instead, Judge Araceli Martinez-Olguin signed off on the settlement that same day, per TechCrunch.
Inside.com.tw, reporting the same approval, put the figure at $1.5 billion (approximately NT$48.5 billion) and noted that the amount sets a record as the largest in the history of U.S. copyright law — a characterization that matches TechCrunch's own description of the deal as "believed to be the largest in the history of U.S. copyright law." Both outlets agree that despite the record size, many authors and creators still do not view the outcome as a clear win, according to TechCrunch.
How Is the Payout Structured, and How Many Works Are Covered?
Per TechCrunch, the payout will deliver $3,000 per work across an estimated 500,000 works, to be shared among the authors and publishers who hold rights to them. Inside.com.tw, citing Reuters, refines this figure: the per-work payment converts to roughly NT$97,000, and the actual number of works covered is closer to 482,000. Of that pool, Inside.com.tw reports that more than 90% — approximately 91% — of works had already been claimed by rights holders as of the approval, with the remainder awaiting disbursement.
| Metric | Figure | Source |
|---|
| Total settlement value | $1.5 billion (~NT$48.5 billion) | TechCrunch; Inside.com.tw |
| Per-work payout | $3,000 (~NT$97,000) | TechCrunch; Inside.com.tw |
| Works covered (TechCrunch estimate) | ~500,000 | TechCrunch |
| Works covered (Reuters figure via Inside.com.tw) | ~482,000 | Inside.com.tw |
| Works already claimed | ~91% | Inside.com.tw |
| Size of pirated book library at issue | 7,000,000+ books | Inside.com.tw |
Where Did Anthropic's Training Data Come From, and What Was Ruled Illegal?
TechCrunch reports that Anthropic built its training library from two sources: books it purchased and scanned, which the court found acceptable, and books it downloaded from pirate sites such as Library Genesis and Pirate Library Mirror. Judge Alsup found the second method illegal on its own terms and said the piracy question could go to trial, according to TechCrunch.
Inside.com.tw adds further detail on the scale of that pirated collection, reporting that Alsup determined Anthropic's long-term retention of a pirated "central library" of more than 7 million books was itself illegal — a finding Inside.com.tw says was sufficient on its own to send that portion of the case to a jury.
What Did Judge Alsup Rule on AI Training and Copyright?
TechCrunch reports that Alsup ruled training an AI model on copyrighted text counts as fair use, a decision the outlet describes as "widely seen as a turning point for the AI industry." However, per TechCrunch, that same ruling drew a line: while training itself could be fair use, the manner in which the underlying data was obtained was a separate legal question — the piracy issue tied to Library Genesis and Pirate Library Mirror downloads.
Inside.com.tw frames the same distinction succinctly, noting that under Alsup's reasoning, training could qualify as fair use, but "how the data was obtained is measured by a different legal yardstick" — a reference to the 7-million-book pirated library finding described above.
Why Won't This Case Set Binding Precedent for the AI Industry?
According to TechCrunch, while the final approval closes out this specific case, it does not settle the legal question industry-wide. That is because Alsup's ruling was a single district court decision, and Anthropic's decision to settle means the case will never reach an appeals court to become binding precedent, TechCrunch reports. In practical terms, other judges remain free to reach different conclusions on fair use in separate AI training lawsuits.
Why Did Anthropic Choose to Settle Instead of Going to Trial?
Inside.com.tw reports that facing the prospect of a potentially enormous jury-determined damages award tied to the piracy findings, Anthropic agreed to settle in September 2025, avoiding trial. As Inside.com.tw puts it, training could still be argued as fair use, but the sourcing of the data was judged by a separate legal standard — the piracy question Alsup had already flagged as trial-worthy.
What Objections Were Raised Before Final Approval, and How Did the Judge Respond?
Per Inside.com.tw, the settlement drew multiple objections before it was approved. Some authors argued the payout amount was still too low; others questioned whether the plaintiffs' legal team's fee cut was excessive; and some claimed certain rights holders were improperly excluded from the payout pool. Judge Martinez-Olguin ultimately rejected these objections and let the settlement stand, according to Inside.com.tw.
Are Other Publishers and Authors Pursuing Similar Copyright Claims Against Tech Companies?
Both TechCrunch and Inside.com.tw report that, just the week before the Anthropic approval, a group of publishers and authors — including Hachette, Cengage, Elsevier, author Scott Turow, and S.C.R.I.B.E. — filed a class-action lawsuit against Google. The suit accuses Google of using their copyrighted works without authorization to train its Gemini AI platform, according to both outlets.
What This Means
The record $1.5 billion figure and the roughly 91% claim rate reported by Inside.com.tw suggest most rights holders in this case are moving toward payment rather than continued litigation. But the same reporting shows the underlying legal question is unresolved: TechCrunch notes Alsup's fair-use ruling was a single district court decision that will never reach an appeals court because Anthropic chose to settle, so it cannot bind other courts. That gap is already visible in practice — the newly filed Hachette-led suit against Google, cited by both TechCrunch and Inside.com.tw, shows publishers and authors pressing the same core question — whether training AI on copyrighted text is fair use — in a fresh forum, even as the Anthropic case closes with payouts rather than a precedent-setting verdict.