IP Law Daily, COPYRIGHT—N.D. Cal.: Landmark Anthropic settlement becomes final, (Jul 21, 2026)

By Matthew Hersh, J.D.

The final judgment approves the settlement as expected, but cuts the requested award of fees.

A settlement agreement that promised to award an expected $3,000 per work to authors and publishers whose books were allegedly used by Anthropic for training of its Claude artificial intelligence model has now been made final, the District Court for the Northern District of California has held. But the court, in rendering a largely-expected opinion that followed the lines of the preliminary settlement and overruled a number of objections to the terms of the agreement, awarded fees and expenses that will likely be a disappointment to the class counsel and named plaintiffs in the action (Bartz v. Anthropic PBC, No. 4:24-cv-05417-AMO (N.D. Cal. Jul. 20, 2026)).

The lawsuit began in August 2024, when authors Andrea Bartz, Charles Graeber and Kirk Wallace Johnson filed a proposed class action alleging that Anthropic copied millions of copyrighted books without permission to train its Claude artificial intelligence models. In June 2025, U.S. District Judge William Alsup ruled on summary judgment that Anthropic's use of lawfully acquired books to train its AI models was protected by fair use, while allowing claims based on the company's alleged acquisition and retention of pirated books to proceed. (Judge Alsup’s ruling, covered by IP Law Daily here, came only days before a colleague on the San Francisco bench federal bench expressed much deeper skepticism about fair use claims in the context of AI.) Before those remaining claims reached trial, the parties negotiated the $1.5 billion class settlement that is now receiving final approval.

Final approval and releases. The court, in a widely expected ruling, granted final approval of the settlement. Applying Rule 23's standards for approval of class action settlements, the court found that notice was properly provided to class members and that the proposed settlement was “fair, reasonable, and adequate.” In reaching this finding, the court noted that there were “no indicia of fraud or collusion” underlying the settlement and that the $1.5 billion fund set aside to compensate authors and publishers provided “meaningful relief” to the class members given the reasonable range of those members’ possible recoveries, “especially since further litigation would likely be complex, expensive, lengthy, and risky.” The court also noted that the estimated per-work payment of approximately $3,000 was “four times the minimum statutory damages amount for willful infringement.”

The court also found that the settlement warranted approval under the remaining factors traditionally considered by courts in the Ninth Circuit. Those factors included, among others, the strength of the plaintiffs’ case, the risk of maintaining class action status throughout the trial, the extent of discovery completed and the stage of the proceedings, the experience and views of counsel, and the reaction of the class members to the proposed settlement. Each of those factors was easily satisfied, the court found, particularly given that the settlement provided “substantial benefits to the [c]lass in light of the novel claims asserted.” Moreover, the court noted, the class’s response was “overwhelmingly favorable,” with the claims rate having reached at least 91.3 percent of the titles at issue—and only 350 valid opt-outs spanning just 1,802 works.

The court also approved the settlement's release provisions, under which participating class members will release claims arising from the conduct at issue in the litigation and the action will be dismissed with prejudice. The court, in language borrowed from its earlier preliminary approval opinion, emphasized that the release does not include any claims related to the enforcement of the settlement (or a related protective order) and also does not include any claims about future conduct. “Under the settlement, class members will give up all claims they could have brought against Anthropic broadly related to past AI inputs, namely pirating and copying their works in the past up to the point of any AI output,” the court noted. “[B]ut they do not give up claims about past AI outputs, nor claims of any kind about future conduct (on or after August 25, 2025).”

Objections. In granting final approval of the settlement, the court also overruled objections filed by class members. Organized by category, the objections challenged, among other things, the reach, content, and scope of the notice; the amount of the settlement fund; the scope of works included in the class; the evidence required to submit valid claims; and the absence of additional forms of relief, such as ongoing licensing arrangements. Rejecting the notice objections, the court pointed to evidence that “506,194 potential Class Members (accounting for 99.5 percent of the works on the Works List) were sent notice not returned as undeliverable”—well within the 70 to 95 percent notice rate that the Federal Judicial Center considers reasonable. The court also rejected the remaining objections, concluding that the settlement provided substantial benefits to the class in light of the risks and uncertainties of continued litigation. Separately, the court ruled on several requests for late opt-outs, denying most but granting two based on excusable neglect.

Attorney fees. Although the court approved the settlement, it declined to award class counsel the full amount of attorney fees they requested. Instead, it awarded approximately $101.6 million in fees, rather than the $187.5 million sought. Although 25 percent of a common fund is generally regarded as a “presumptively reasonable” benchmark, the court concluded that a lodestar analysis, rather than a percentage-of-the-fund approach, was more appropriate in a “megafund” case such as this one. Awarding 12.5 percent of the settlement fund, the court found, would have produced a 6.92 multiplier on class counsel's lodestar—“far outside the range of multipliers found to be reasonable in mega fund cases within this Circuit.” Instead, the court applied a 3.77 multiplier. The court also substantially reduced the requested service awards for the named plaintiffs, approving payments of $15,000 each instead of the $50,000 requested.

The Case is No. 4:24-cv-05417-AMO.

Judge: Martínez-Olguín, A.

Attorneys: Justin A. Nelson (Susman Godfrey L.L.P.) for Andrea Bartz. Douglas Andrew Winthrop (Arnold & Porter Kaye Scholer LLP) for Anthropic PBC.

Companies: Anthropic PBC

MainStory: TopStory Copyright CaliforniaNews AINews GCNNews