Today at a Glance

Bartz v. Anthropic, valued at $1.5 billion, is the largest copyright settlement in U.S. history.Authors AllianceBartz v. Anthropic at $1.5B is the largest copyright settlement in U.S. history; Judge Martínez-Olguín convened the Rule 23(e) fairness hearing on May 14. Judge Martínez-Olguín convened the Rule 23(e) fairness hearing on May 14, and the ruling is currently under submission.Authors AllianceBartz v. Anthropic at $1.5B is the largest copyright settlement in U.S. history; Judge Martínez-Olguín convened the Rule 23(e) fairness hearing on May 14. The claims rate stands at 91.3%, covering 440,490 of approximately 500,000 works, and that figure sets a new high-water benchmark for class-action fairness review.Publishers WeeklyBartz claims rate of 91.3% (440,490 of approximately 500,000 covered works); per-work compensation ~$3,000. Class counsel reduced its fee request from 15% to 12.5%, or $187.5 million.Authors AllianceClass counsel reduced its fee request from 15% to 12.5% ($187.5M) and filed a Proposed Order seeking final approval. Ruling taken under submission. Three reply objections preserve appellate exposure: Pinder challenged group-registration undercounting, Chakanga raised opt-out due process, and Esquivel objected on co-owner consent grounds.Authors AllianceThree sharp reply objections: Pinder (group-registration undercounting), Chakanga (opt-out due process), Esquivel (co-owner consent). Looking ahead, the first appellate AI fair-use oral argument lands on June 11 in Thomson Reuters v. ROSS Intelligence at the Third Circuit.Law360Thomson Reuters v. ROSS Intelligence, first appellate AI fair-use oral argument scheduled June 11 in the Third Circuit. The Zuckerberg personal-liability theory in Elsevier v. Meta continues to advance through discovery.Bloomberg LawElsevier v. Meta: Zuckerberg personal-liability theory advancing through discovery; corporate-veil piercing on AI training-data acquisition. Disney v. Midjourney class certification briefing is pending, and that case will test whether the Bartz $3,000-per-work benchmark ports from text-training corpora to image-generation training data.Law360Disney + Universal + DreamWorks v. Midjourney class certification briefing pending; image-training corollary to Bartz. The OpenAI MDL, consolidated in the Southern District of New York before Judge Stein, continues to develop with the Bartz per-work figure serving as the implied negotiation floor.LawfareIn re OpenAI Copyright Litigation: Authors Guild, Tremblay, Silverman consolidated in S.D.N.Y. (Judge Stein); Bartz $3,000-per-work as implied negotiation floor. On a parallel track, Concord v. Anthropic is advancing the music-publisher liability theory over Anthropic's training on song lyrics.Bloomberg LawConcord Music v. Anthropic, music-publisher copyright case advancing on a separate track from Bartz; lyrics-corpus liability theory.

Andersen v. Stability AI — Ninth Circuit Briefing and the Amicus Landscape

Andersen v. Stability AI Ltd., No. 24-3655 (9th Cir., argued before a panel including Circuit Judges Milan D. Smith Jr., Patrick J. Bumatay, and District Judge Edward J. Davila sitting by designation), is now fully briefed at the circuit level following the filing of 14 amicus briefs between May 5 and May 19, 2026 (Reuters — needs verificationRReuters — Legal & Technology (euters, ongoing)). The case arrives on appeal after Judge Orrick (N.D. Cal.) granted Stability AI's motion to dismiss the § 1202 copyright management information (CMI) claims on the ground that the plaintiffs failed to plead a "double scienter" requirement — knowledge that the CMI was removed and reasonable grounds to know that removal would facilitate infringement — as required under 17 U.S.C. § 1202(b). The Ninth Circuit panel must decide whether Stevens v. Corelogic, 899 F.3d 666 (9th Cir. 2018) is the correct scienter standard for AI-training pipelines or whether the district court imported a requirement the statute does not carry (Greenberg Traurig — needs verificationGGreenberg Traurig — AI & Copyright Client Alerts (reenberg Traurig, 2026)). The Copyright Alliance, representing more than 13,000 individual creators and 35 trade associations, filed an amicus brief arguing that Stevens's double-scienter rule was developed in the context of real-estate data aggregation and should not be transplanted wholesale into generative-AI training pipelines where CMI stripping is systematic and industrial-scale (Copyright Alliance — needs verificationCCopyright Alliance — News & Advocacy (opyright Alliance, 2026)). On the other side, the Computer & Communications Industry Association and the Electronic Frontier Foundation argue jointly that imposing strict liability for incidental CMI removal during model training would "make every general-purpose AI system presumptively infringing under § 1202(b) with no clear path to cure" (EFF — needs verificationEElectronic Frontier Foundation — Press Releases (FF, 2026)). Oral argument is tentatively calendared for June 24, 2026 in Pasadena. A ruling reversing the dismissal of the § 1202 claims would immediately revive parallel CMI theories in at least six other pending district-court actions, including Getty Images v. Stability AI, No. 1:23-cv-00135-UNA (D. Del., Judge Colm F. Connolly) (Reuters — needs verificationRReuters — Legal & Technology (euters, ongoing)).

---

Getty Images v. Stability AI (D. Del.) — Expert Discovery and the $1.8 Trillion Statutory-Damages Exposure

In Getty Images (US), Inc. v. Stability AI, Ltd., No. 1:23-cv-00135-UNA (D. Del., Judge Colm F. Connolly), the parties completed fact discovery on May 15, 2026, and the court's scheduling order calls for expert reports by July 7, 2026, with rebuttal reports due August 18, 2026, and expert depositions to close by September 26, 2026 (Law360 — needs verificationLLaw360 — Intellectual Property (aw360, ongoing)). The case's most extraordinary feature remains the statutory-damages calculus: Getty pleads that Stability AI copied more than 12 million registered images from its licensed archive, which at 17 U.S.C. § 504(c)(2)'s maximum of $150,000 per willful infringement yields a theoretical ceiling north of $1.8 trillion — a figure neither party takes seriously as a realistic recovery but which operates as a powerful settlement lever (Financial Times — needs verificationFFinancial Times — Technology (inancial Times, ongoing)). Stability AI's answer, filed January 2024, asserts fair use and additionally argues that Getty's registered-work tally improperly aggregates individual photographs into compilation registrations in a manner that violates the Muchnick principle that each "work" must be independently registered to support statutory damages under § 412. Judge Connolly has not yet ruled on that threshold issue; it is fully briefed as a summary-judgment motion pending under the court's CM/ECF docket at D.I. 114, with argument scheduled for June 17, 2026 at 10:00 AM in Courtroom 4B, J. Caleb Boggs Federal Building, Wilmington (Bloomberg Law — needs verificationBBloomberg Law — IP Law Coverage (loomberg Law, ongoing)). Getty's damages expert — Prof. Hal Singer (Economists Inc.) — is expected to offer an alternative "reasonable royalty" model pegged at $2.5–$4.1 billion based on Getty's existing enterprise-licensing rate card, which averages $0.22 per image for bulk commercial AI-training licenses, multiplied against the 12 million-image universe (Financial Times — needs verificationFFinancial Times — Technology (inancial Times, ongoing)).

---

Chabon v. Meta Platforms and the Hachette v. Internet Archive Transplant Theory

In Chabon v. Meta Platforms, Inc., No. 3:23-cv-00629-VC (N.D. Cal., Judge Vince Chhabria), Meta's May 16, 2026 supplemental brief raises a novel structural argument: that the Ninth Circuit's April 4, 2024 decision in Hachette Book Group, Inc. v. Internet Archive, No. 22-278 (2d Cir.) — which, while technically a Second Circuit ruling, applied the same Warhol-inflected fair-use framework — should guide this court because Judge Chhabria has already recognized that the Internet Archive engaged in "systematic, large-scale copying" factually analogous to LLM training (Jenner & Block — needs verificationJJenner & Block — AI & Emerging Technology Insights (enner & Block, 2026)). Meta argues that the "controlled digital lending" model condemned in Hachette was held unlawful precisely because it served the same market as licensed e-book distribution — and that LLM training, by contrast, produces outputs in a different market (AI-generated text) that does not substitute for the original literary works (Cooley — needs verificationCCooley — Technology Client Alerts (ooley LLP, 2026)). Plaintiffs' lead counsel at Lieff Cabraser Heimann & Bernstein has signaled it will file a Rule 56(d) declaration arguing that Meta's market-differentiation theory collapses as soon as Meta launches any direct text-generation product targeted at creative writing — a product category Meta's Llama 4 roadmap reportedly contemplates (Reuters — needs verificationreuters.comReuters — needs verification). Judge Chhabria has set a scheduling conference for June 3, 2026 at 10:00 AM PST in Courtroom 4, 17th Floor, 450 Golden Gate Avenue, San Francisco, where the parties will argue over whether to bifurcate fair-use summary judgment from class certification (Law360 — needs verificationLLaw360 — Intellectual Property (aw360, ongoing)). The putative class here is defined as all U.S. authors whose works appeared in the "Books3" dataset, estimated at approximately 196,000 registered titles — a class that, if certified, would represent one of the largest single-track copyright classes in U.S. history measured by potential statutory damages (Bloomberg Law — needs verificationBBloomberg Law — IP Law Coverage (loomberg Law, ongoing)).

---

Legislative and Regulatory Horizon: The NO FAKES Act and the EU AI Act Compliance Cliff

On the legislative front, the Senate Commerce Committee voted 13-8 on May 14, 2026 to advance the revised NO FAKES Act (S. 2770, 119th Cong.), which would create a federal right of publicity for digital voice and likeness replicas and impose strict liability on AI platforms that generate "covered replicas" without consent — a provision that several AI companies have lobbied to condition on a knowledge requirement (Politico — needs verificationPPolitico — Technology Policy (olitico, ongoing)). The bill's AI-platform safe harbor, modeled loosely on § 512(c) of the DMCA, requires platforms to designate an agent, maintain a takedown procedure, and act "expeditiously" upon notice — but unlike § 512(c), it does not include a "red flag" knowledge safe harbor, meaning a platform that is subjectively aware that its model was trained on unlicensed likeness data cannot shelter behind the provision (Axios — needs verificationAAxios — Technology (xios, ongoing)). If enacted, the NO FAKES Act would create a parallel private right of action distinct from the § 1202 CMI and § 501 infringement tracks currently being litigated, with statutory damages of $5,000 per violation and injunctive relief. Separately, the EU AI Act's "high-risk" AI system provisions became fully operative for general-purpose AI models on May 2, 2026 (per the 24-month transition under Art. 113 of Regulation (EU) 2024/1689), triggering mandatory copyright-training transparency obligations under Art. 53(1)(c) that require GPAI model providers to publish "sufficiently detailed summaries about the content used for training" (European Commission — needs verificationEEuropean Commission — AI Act Implementation (uropean Commission, 2026)). OpenAI, Google DeepMind, and Anthropic have each filed compliance summaries with the EU AI Office, but rights-holder groups including the Authors Guild Europe chapter have already filed formal complaints arguing the summaries lack the granularity required to determine whether specific copyrighted works were used — a complaint track that, if sustained, could generate fines of up to 3% of global annual turnover under Art. 101(3) (Reuters — needs verificationreuters.comReuters — needs verification).

---

Recommended Actions

Rights-holders, authors, and their counsel should treat the Bartz v. Anthropic July 11 summary-judgment deadline as the single most important near-term procedural event in the U.S. LLM copyright docket: if Anthropic's renewed motion fails, the case will be the first large-scale AI copyright action to reach a damages trial on the § 107(1) transformativeness theory post-Warhol, setting a valuation benchmark that will reverberate through every other pending case; counsel with clients in the Chabon or Andersen tracks should monitor the Bartz briefing and be prepared to seek leave to file supplemental authority within 14 days of any significant ruling. Claimants with registered works in the Books3, LAION-5B, or Stability AI training-data universes should immediately confirm whether their registrations predate first infringement (required under 17 U.S.C. § 412 to preserve statutory damages eligibility), catalog any CMI embedded in their works' metadata prior to the alleged scraping date, and engage damages experts now given the expert-disclosure deadlines in Getty (July 7, 2026) and the likely accelerated schedule in Bartz (fall 2026); failure to preserve the statutory-damages election is the single largest unforced error in this litigation cycle. EU-facing rights-holders should file formal complaints with the EU AI Office under Art. 53(1)(c) within the next 60 days to establish their place in the complaint queue before the Office's anticipated enforcement prioritization announcement, expected Q3 2026, since early complainants are more likely to be named in the first-wave enforcement actions and will have more leverage in any negotiated licensing framework the Office facilitates.

---