Yesterday’s Rule 23(e) fairness hearing in Bartz v. Anthropic PBC, 3:24-cv-05417 (N.D. Cal.), convened at 2:00 p.m. PT before Judge Araceli Martínez-Olguín in Courtroom 12 of the San Francisco Federal Courthouse, with Zoom access available to the public. Class Counsel submitted a Proposed Order (Dkt. 646-1) requesting final approval of the $1.5 billion settlement — the largest copyright resolution in United States history — and overruling all objections, while simultaneously disclosing a downward revision of the fee request from 15 percent to 12.5 percent of the gross fund, yielding a petition of approximately $187.5 million plus $3 million in incurred expenses, an $18.22 million cost reserve, and $50,000 service awards to each of the three class representatives. The claims data remains extraordinary: 440,490 of 482,460 eligible works have been claimed, a 91.3 percent rate reported in Class Counsel’s April 16 updated claims filing (Dkt. 643), though Authors Alliance correctly notes that because large publishers holding rights in tens of thousands of works presumably filed early, the percentage of individual authors who responded is likely more modest than the headline figure suggests. The court did not issue a ruling from the bench; the matter was taken under submission, meaning the final-approval order could issue at any time in the coming days or weeks. (Authors Alliance)

Several reply objections filed on the eve of the hearing sharpen the structural challenges the court must now address. Victoria Pinder (Dkt. 648) renewed her group-registration objection, documenting that her 40-plus independently published novels registered under a single lawful group copyright number count as one “claimable Work,” meaning an identically prolific author who registered individually would receive forty shares to her one — a disparity Class Counsel’s Proposed Order did not engage on the merits. Chinonto Chakanga (Dkt. 649) challenged the adequacy of the opt-out process on due-process grounds, arguing that thousands of pages of sealed summary-judgment records were not unsealed until January 21, 2026, eight days before the original opt-out deadline, and citing the Ninth Circuit’s January 28 decision in Avery v. TEKsystems for the proposition that compressed timing combined with a sealed judicial record is a substantive defect. Laura Esquivel (Dkt. 651), the Mexican author of Como agua para chocolate, contested Class Counsel’s novel requirement that she obtain co-owner consent for her late opt-out, noting that the Long Form Notice expressly provides that a single rightsholder’s opt-out decision binds co-authors and publishers, and argued excusable neglect under the Pioneer standard given that she received no Spanish-language notice and suffered a stroke during the relevant period. The James Sills objection (Dkt. 600) — reporting that Elsevier proposed an 88/12 publisher-author split for his textbook — remains a focal point for Professor Lea Bishop’s intra-class fairness theory (Dkt. 602). (Authors Alliance)

On the parallel docket front, the Elsevier Inc. et al. v. Meta Platforms, Inc. class action filed May 5 in S.D.N.Y. continues to draw scrutiny for its personal-liability theory against Mark Zuckerberg. The complaint alleges Zuckerberg personally authorized and actively encouraged the torrenting of 267 terabytes of pirated material from LibGen and Anna’s Archive to train Meta’s Llama models, after Meta abandoned a $200 million licensing budget at Zuckerberg’s instruction, and five major publishing houses — Hachette, Macmillan, McGraw Hill, Elsevier, and Cengage — together with Scott Turow seek statutory damages, a permanent injunction, and destruction of all infringing copies. Authors Alliance published a pointed commentary questioning whether publishers are purporting to represent author interests without author consent. Meta has vowed to fight the lawsuit aggressively. (AAP) (Variety)

The appellate landscape is crystallizing around the June 11 oral argument in Thomson Reuters Enterprise Centre GmbH v. Ross Intelligence Inc. (3d Cir. No. 25-2153), which will be the first appellate fair-use argument on AI training in any United States court. Supplemental briefs filed May 11 reveal sharply divergent readings of the Third Circuit’s recent ASTM v. UpCodes ruling: ROSS Intelligence argues the ruling demands summary reversal, while Thomson Reuters counters that ROSS’s purpose was commercially substitutive rather than access-oriented. The panel’s treatment of the objective-versus-subjective transformative-purpose distinction will directly inform fair-use analysis across every pending AI copyright docket, including the Concord/UMG v. Anthropic music-publishing litigation where the cross-MSJ hearing is set for July 15 before Judge Lee. (XIRA/LawNext) (Music Business Worldwide)

On the legislative front, the NO FAKES Act and TRAIN Act continue to build bipartisan momentum following the Blackburn-Welch roundtable with over twenty artists on April 22. The Generative AI Copyright Disclosure Act (Dean-Moran) and the CLEAR Act (Schiff-Curtis) would require AI developers to disclose copyrighted training data, while the TRUMP AMERICA AI Act discussion draft would amend 17 U.S.C. § 107 to declare that AI training on copyrighted works does not qualify as fair use — creating direct tension with the judicial determinations pending in Bartz, Thomson Reuters, and the OpenAI MDL. The Supreme Court’s March 2 denial of certiorari in the Thaler AI-authorship case reaffirmed human authorship as foundational to U.S. copyright. The licensing-deal landscape continues to serve as a comparator for settlement valuation, with the Anthropic $1.5 billion settlement, the NYT-Amazon deal at $20–25 million, and the Google-Reddit agreement establishing the pricing matrix. (Senate.gov/Blackburn) (AI Business)

Class members with claims already filed should verify their payment information through the settlement portal at anthropiccopyrightsettlement.com in advance of the June 11 distribution computation, and authors who registered works under group copyright registrations should prepare supplemental documentation identifying each distinct title, as the Pinder reply objection (Dkt. 648) and the court’s eventual treatment of the group-registration undercounting methodology will determine whether additional per-title claims become available. Rights-holders who opted out should monitor the docket closely for Judge Martínez-Olguín’s final-approval order, which could issue at any time, as the court’s treatment of the per-work valuation framework, the publisher-author allocation methodology, and the Esquivel co-owner-consent theory will directly inform litigation-value assessments and opt-out strategies in parallel dockets. Music publishers with compositions in Anthropic’s training corpus should coordinate with RIAA and NMPA counsel on the Concord cross-MSJ briefing ahead of the July 15 hearing. AI developers should conduct outbound-training-data audits against both the Bartz and Elsevier v. Meta frameworks, track the Thomson Reuters v. ROSS oral argument on June 11 for the Third Circuit’s treatment of the transformative-purpose test, and evaluate DMCA § 1202 CMI-stripping exposure. All stakeholders should track the legislative calendar, as the TRUMP AMERICA AI Act’s proposed statutory override of fair use for AI training would fundamentally alter the risk calculus in every pending docket.