Anthropic has agreed to pay $1.5 billion to resolve a copyright lawsuit brought by book authors, and a federal judge granted final approval in July 2026. The settlement resolves claims involving allegedly pirated copies of books acquired from repositories including Library Genesis and Pirate Library Mirror. It does not establish that using lawfully acquired copyrighted books to train an AI model is generally illegal.
The ordinary claim and opt-out deadlines have passed. The case’s lasting importance is the distinction between how Anthropic acquired books and how it used books to train Claude.
| # | Preview | Product | Price | |
|---|---|---|---|---|
| 1 |
|
Copyright Law | $144.88 | Buy on Amazon |
| 2 |
|
Copyright Law: Cases and Materials (v8.0) | $19.87 | Buy on Amazon |
| 3 |
|
Copyright Law of the United States: and Related Laws Contained in Title 17 of the United States Code | $10.32 | Buy on Amazon |
| 4 |
|
Copyright Law in a Nutshell | $65.00 | Buy on Amazon |
| 5 |
|
Copyright Handbook, The: What Every Writer Needs to Know | $37.99 | Buy on Amazon |
As an Amazon Associate I earn from qualifying purchases.
The short version
- Settlement amount: $1.5 billion.
- Who sued: Authors including Andrea Bartz, Charles Graeber, and Kirk Wallace.
- What was allegedly copied: Books obtained from pirate repositories, principally Library Genesis (LibGen) and Pirate Library Mirror (PiLiMi).
- What the court previously decided: Anthropic’s use of books to train its models qualified as fair use under the circumstances addressed in the June 2025 ruling.
- What remained unresolved: Whether downloading, copying, and retaining books from pirate sources created separate copyright liability.
- Can new ordinary claims be filed? No. The settlement website lists March 30, 2026 as the claim deadline and February 9, 2026 as the opt-out and objection deadline.
The settlement is a negotiated resolution, not a nationwide rule that AI training is lawful or unlawful. It also is not an admission that Anthropic violated every copyright theory asserted by the authors.
Free tools Windows power users keep installed
One-click scans. No signup required.
What was Bartz v. Anthropic PBC?
The case, formally Bartz et al. v. Anthropic PBC, No. 3:24-cv-05417, was filed in the U.S. District Court for the Northern District of California on August 19, 2024. The authors alleged that Anthropic used copyrighted books without permission in developing Claude, its family of AI models.
#1 Best Overall
The dispute became more specific than the question of whether copyrighted books appeared in training data. The authors focused on Anthropic’s alleged acquisition and storage of books from unauthorized sources. Settlement materials identify LibGen and PiLiMi as the relevant repositories and say Anthropic downloaded approximately 7 million files from those sources.
That figure should not be read as 7 million distinct qualifying books, nor does the existence of a book in one of those repositories automatically make it eligible for payment. The approved Works List, created from materials produced in the litigation, controls eligibility.
The crucial distinction: training versus acquiring pirated copies
In a June 23, 2025 ruling, Judge William Alsup treated Anthropic’s use of copyrighted books to train its models as fair use under the facts presented to the court. That was an important ruling for Anthropic, but it did not end the case.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
The court did not grant Anthropic summary judgment on claims concerning books allegedly obtained from pirate libraries. Those claims involved the acquisition, copying, and retention of unauthorized source files. They remained potential infringement issues that could have proceeded to trial.
In practical terms, the ruling did not say either of the following:
- “All AI training on copyrighted books is illegal.”
- “Everything an AI company does with copyrighted books is lawful if the books are used for training.”
Instead, it separated two legal questions:
- Use in model training: The court found this use to be fair use in the circumstances before it.
- Source-copy acquisition and retention: The alleged use of pirate repositories created separate unresolved risks.
That distinction helps explain why Anthropic could receive a favorable fair-use ruling and still agree to a very large settlement.
Why did Anthropic settle?
Settlement avoided a trial over the unresolved piracy-related claims, along with the cost, uncertainty, and possible appeals that would have followed. The potential exposure was also significant because copyright law can provide statutory damages in appropriate circumstances.
Rank #2
Anthropic preserved its favorable ruling on the training issue while avoiding a full trial on whether its acquisition and retention of allegedly pirated books infringed the authors’ rights. The authors, meanwhile, obtained a substantial recovery without having to prove every claim through trial and appeal.
The agreement should not be described as a judicial finding that Anthropic “stole” every book used to train Claude. The more precise description is that the authors alleged Anthropic acquired books from pirate sources, and Anthropic settled the covered claims without a final trial judgment resolving all of those allegations.
What does the $1.5 billion settlement cover?
The settlement fund
The gross settlement fund is $1.5 billion. Published coverage has reported an approximate recovery of about $3,000 per qualifying book or work, but that is not a guaranteed payment to every author or rights holder.
Actual distributions depend on the approved allocation formula, the number of valid claims, whether a work qualifies, ownership interests, and the information accepted by the settlement administrator. The $3,000 figure is best understood as a reported estimate, not a fixed price that Anthropic agreed to pay for every book.
The Tool Desk
Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Legal fees and costs
The court-approved attorneys’ fees and litigation costs were reported by the Authors Guild at approximately $101.56 million, or about 6.8% of the fund. The final order reportedly withheld 10% of the fee award pending a later accounting.
Covered conduct
The release covers defined claims concerning Anthropic’s past acquisition, copying, retention, and use of works obtained from the specified pirate datasets through August 25, 2025.
The settlement does not generally release:
- Claims involving conduct after August 25, 2025;
- Claims concerning allegedly infringing outputs from Anthropic models;
- Claims involving works outside the approved class and Works List; or
- Claims preserved by people who properly opted out before the deadline.
Destruction obligations
The settlement includes obligations concerning original files downloaded or torrented from LibGen and PiLiMi, and copies originating from those sources, subject to legal-preservation requirements. This is a requirement concerning specified files and datasets.
Rank #3
It should not be described as an order to erase Claude’s “knowledge” or as proof that trained model weights can simply be purged. Destroying source files and removing data from a trained model are separate technical questions, and the settlement materials do not establish that Anthropic erased all information that might be reflected in model behavior.
Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minutePC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Who could claim money?
Eligibility generally required a claimant to be the legal owner, beneficial owner, or sole owner of the exclusive right to reproduce a qualifying copyrighted book. The book also had to appear on the official Works List.
This distinction matters because authors and publishers may hold different interests. A publisher may own the exclusive reproduction right while an author receives royalties as a beneficial owner. Publishing agreements, rights reversions, co-author arrangements, work-for-hire provisions, and other contracts can affect who is entitled to recover and how any payment is divided.
Being an author whose book appeared online, or whose book was found somewhere in LibGen or PiLiMi, was not enough by itself. The settlement’s Works List was the relevant reference point.
The settlement website also cautioned that editions and formats may have different ISBNs, ASINs, copyright registrations, authors, or content. A metadata mismatch did not necessarily defeat a legitimate claim, but the claimant still needed a valid ownership basis.
Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Important dates—and what they mean now
| Date | Event |
|---|---|
| August 19, 2024 | The case was filed. |
| June 23, 2025 | The court issued its mixed fair-use ruling. |
| August 26, 2025 | The parties announced a settlement in principle. |
| February 9, 2026 | Deadline to opt out or object. |
| March 9, 2026 | Deadline to rejoin after opting out. |
| March 30, 2026 | Deadline to submit a claim. |
| May 14, 2026 | Final-approval hearing. |
| July 2026 | Final approval and judgment were entered. |
As of the current settlement status, the ordinary claim and opt-out periods have passed. Anyone who believes they missed the deadline should not assume that a late claim will be accepted. The official settlement website and court documents are the appropriate sources for any administrator- or court-authorized procedure.
Does this mean AI training on copyrighted books is illegal?
No. The settlement does not decide that question for every AI company, every book, or every method of acquiring data.
Rank #4
Judge Alsup’s 2025 ruling treated Anthropic’s training use as fair use under the specific facts before the court. But the case settled before the unresolved piracy-related claims reached a full trial and appellate review. The settlement therefore does not create nationwide appellate precedent or a universal license for AI companies to copy books.
It also does not establish that every AI company must license every copyrighted book before training a model. Other cases may involve different sources, contracts, copying practices, model architectures, outputs, or facts about market harm.
What the case means for authors, publishers, and AI companies
The practical lesson is about data provenance as much as training itself. Companies developing AI systems face different legal and business risks depending on:
- Where the source material came from;
- Whether the company lawfully acquired the copies;
- Whether it maintained an unauthorized library of works;
- Whether the copies were used for training, research, storage, or another purpose; and
- Whether later conduct falls within a particular settlement release.
For authors and publishers, the settlement may strengthen the argument that unauthorized acquisition and large-scale retention of copyrighted works deserve separate scrutiny from the question of transformative model training. For AI companies, it highlights the litigation risk created by questionable source datasets even when a court may view the eventual training use favorably.
The dollar amount may influence future negotiations, but it is not a statutory tariff or a universal per-book price for lawful AI training. This was a class-specific settlement tied to defined works, claims, and allocation rules.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Common misconceptions
“Anthropic paid because AI training on copyrighted books is illegal.”
That is incomplete. The court treated the training use as fair use while leaving claims involving allegedly pirated copies unresolved.
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Fix the driver behind crashes, sound loss and screen glitches3Repair Windows errors before they cause bigger problems“Every author gets $3,000.”
No. About $3,000 was reported as an approximate per-work figure. Payment depended on valid claims, qualifying works, ownership, and the allocation process.
Best Value
“Every book used by Claude was proven to be pirated.”
No. The settlement concerns specified datasets and the approved Works List. It does not establish that every book used in training came from an unauthorized source.
“The settlement bans AI training on books.”
No. It resolves defined claims against Anthropic and does not create a general ban.
“Destroying the files means Claude forgot the books.”
No. The destruction obligations concern specified source files and copies. They do not establish that trained model weights or model behavior were fully purged.
Recommended Free Tools
“This is the final word on AI copyright.”
No. The settlement is important, but it is not a nationwide appellate ruling. Other disputes involving books, journalism, images, music, and AI outputs may raise different legal questions.
What happens next?
The relevant next phase is settlement administration and distribution, not a new general claim period. The final approval resolves the covered litigation, while the administrator handles valid claims under the court-approved process.
For the wider AI industry, the case is likely to remain influential because it demonstrates that “training” and “data acquisition” cannot always be analyzed as one undifferentiated activity. But its most important legal questions remain fact-specific, and the settlement did not supply a universal answer for all copyrighted works or all AI systems.
Quick Recap
Sources
- Official settlement website
- Settlement documents
- Claim-form eligibility rules
- Settlement dates
- Federal court case page
- Authors Guild account of final approval
- Associated Press coverage of final approval
- Associated Press coverage of the 2025 fair-use ruling
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
