> ## Content Index > Fetch the complete content index at: https://dispatchdaily.media/llms.txt > Use this file to discover other available public pages before exploring further. # Sarah vs. the Machines - URL: https://dispatchdaily.media/untitled-2/ - Published: 2026-08-18T19:01:00.000Z - Updated: 2026-09-03T03:11:30.000Z - Description: The copyright war is entering its second act: training, acquisition and substitution are becoming three different legal battles. - Author: Joe Bel Bruno - Tags: #hero, #copyright Copyright Dispatch Intelligence / AI Rights / 02 September 2026 # Sarah *vs.* the Machines The copyright war is entering its second act. The question is no longer simply whether AI can learn from copyrighted work. It is who controls the corpus, who can prove economic harm—and who gets paid. Premium intelligence12-minute readCase status: active The Dispatch View ## AI companies are winning the argument that machines can learn from copyrighted work. They are not winning the argument that they can steal the library first. 01 **Training**Increasingly protected → 02 **Acquisition**The new battleground → 03 **Output / substitution**Still unresolved 01 ## The caseboard The industry is not fighting one copyright question. It is fighting three—and the cases are splitting along those lines. Matter Fair use Provenance Market harm **OpenAI**Silverman / NYT / Authors Guild Open Contested Unproven **Meta**Kadrey Developer win Disputed Record-specific **Anthropic**Bartz Training protected Pirate library Not reached **Music**Publishers v. Anthropic New front Alleged piracy Testing favors defendantrisk / adverseunresolved **$1.5B**Anthropic authors’ settlement, approved July 2026 **12M**Getty images alleged in the Stability AI litigation **$203M**Shutterstock 2025 data, distribution and services revenue **$140M**Reddit 2025 “other revenue,” including content licensing 02 ## Copyright is becoming two businesses ### Liability - Piracy and unauthorized copying - Litigation and statutory damages - Regurgitation and substitution - Unpriced legacy exposure PROVENANCE ### Asset - Clean, licensable corpora - Rights and ownership metadata - Authenticated access - Freshness, scarcity and indemnity The real story ## The battlefield moved Sarah Silverman sued OpenAI after alleging that her memoir, *The Bedwetter*, had been copied without authorization into training datasets. Three years later, that question almost looks quaint. Her case now sits inside a consolidated New York fight involving authors, publishers and *The New York Times*. What began as a claim about one writer’s book has become a forensic examination of how the modern AI corpus was built. Two federal rulings supplied the industry’s emerging doctrine. In *Bartz v. Anthropic*, Judge William Alsup treated model training as transformative fair use on the record before him—but separated that use from the creation of a permanent library assembled with pirated books. In *Kadrey v. Meta*, Judge Vince Chhabria ruled for Meta on a thin record while emphasizing the plaintiffs’ failure to show meaningful market dilution. **Dispatch analysis:** Training, acquisition and output are becoming distinct legal products. A developer may have a strong defense for what a model learned and still face enormous exposure for how the source material entered the pipeline. ### The $1.5 billion warning Anthropic’s settlement turns provenance from an abstract compliance concern into a line item. The lesson is not that every training use requires a royalty. It is that a fair-use destination does not necessarily cleanse an infringing route. The acquisition record—purchase, license, scrape or pirate archive—may determine the bill. ### Where the money is moving The market is already rewarding clean access. Shutterstock’s Data, Distribution and Services revenue reached $203.3 million in 2025, up from $137.3 million in 2023\. Reddit reported $140 million of “other revenue” in 2025, a category that includes content licensing but is not broken out separately. These are not standardized royalties. They are early evidence of a wholesale market for legally usable human knowledge. **Dispatch analysis:** The premium asset is becoming copyright plus access plus provenance plus machine-readable rights. Large catalogs and data intermediaries gain leverage; individual creators risk retaining rights without negotiating scale. Live case file / active ### IN RE OPENAI COPYRIGHT INFRINGEMENT LITIGATION Venue S.D.N.Y. Plaintiffs Authors · publishers · news organizations Core test Training · acquisition · substitution Gov’t view Fair-use intervention supports OpenAI Next signal How the court separates corpus provenance from model use Dispatch risk ●●●○ Elevated 03 ## The five-second liability test The path matters as much as the destination. A copyrighted work ↓ How did you get it? **Bought / licensed** Documented access and rights **Pirated / disputed** Permanent-library exposure ↓ Model training ↓ What did the model do? **Learned / transformed** Fair-use argument strengthens **Reproduced / substituted** Market-harm argument strengthens That path may determine liability. 04 ## Who’s winning? **AI developers** ▲ Training doctrine is moving their direction—on specific records. **Major rights owners** ▲ Clean catalogs are becoming strategic, defensible inputs. **Data intermediaries** ▲▲ Rights verification and corpus cleansing become infrastructure. **Individual creators** ? Rights remain; negotiating scale and proof of harm lag. **Pirated corpus builders** ▼▼ The increasingly dangerous link in the training-data chain. Dispatch terminal / monitoring ## What Dispatch is watching 01 Does another court clearly separate **training** from **acquisition**? 02 Does a court finally quantify AI-driven **market substitution**? 03 Does music establish a different licensing precedent because its rights systems are already institutionalized? 04 Do clean training datasets begin appearing as material, separately disclosed corporate revenue? 05 Does provenance become standard diligence for model developers, investors and insurers? 05 The signal we’re building ## Dispatch AI Rights Ledger A continuously maintained record of the legal, licensing and provenance exposure embedded in major AI models. | Model / company | Corpus | Source status | Rights status | Lead matter | Exposure | | --------------- | --------------------- | ---------------------------- | --------------- | ------------ | -------- | | OpenAI | Books1 / Books2; news | Disputed | Mixed / unknown | S.D.N.Y. MDL | ●●●○ | | Anthropic | Books | Purchased + pirate libraries | Mixed | Bartz | ●●●● | | Meta | Books | Disputed | Disputed | Kadrey | ●●○○ | | Stability AI | Images | Web-scraped / alleged | Contested | Getty | ●●●○ | Prototype / AI Copyright Risk Index 68 / 100 · elevated Illustrative framework, not an investment or legal-risk rating. Proposed inputs: corpus source confidence, rights coverage, active claims, adverse orders, output similarity, market-substitution evidence, indemnity and licensed-data share. 06 ## Primary record **Bartz v. Anthropic:** [June 2025 fair-use order](https://assets.fenwick.com/documents/Bartz-v.-Anthropic-Fair-Use-Opinion.pdf?ref=dispatchdaily.media). **Kadrey v. Meta:** [June 2025 summary-judgment order](https://bpb-us-e2.wpmucdn.com/sites.uci.edu/dist/d/2220/files/2025/07/Kadrey-v-Meta-Platforms-Inc%5FRedacted.pdf?ref=dispatchdaily.media). **U.S. Copyright Office:** [Copyright and Artificial Intelligence, Part 3](https://www.copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-3-Generative-AI-Training-Report-Pre-Publication-Version.pdf?ref=dispatchdaily.media). **Shutterstock:** [2025 results](https://investor.shutterstock.com/news-releases/news-release-details/shutterstock-reports-full-year-2025-and-fourth-quarter-financial?ref=dispatchdaily.media). **Reddit:** [2025 Form 10-K](https://www.sec.gov/Archives/edgar/data/1713445/000171344526000022/rddt-20251231.htm?ref=dispatchdaily.media). **Getty v. Stability AI:** [U.S. complaint](https://www.docketalarm.com/cases/Delaware%5FDistrict%5FCourt/1--23-cv-00135/Getty%5FImages%5F%28US%29%5FInc.%5Fv.%5FStability%5FAI%5FInc/1/?ref=dispatchdaily.media). **OpenAI litigation:** [OpenAI filing index and case statements](https://openai.com/new-york-times/?ref=dispatchdaily.media). Dispatch analysis separates reported fact from inference. Litigation outcomes remain record-specific and subject to appeal. Last reviewed 02 September 2026.