# OpenAI and Microsoft Clash with News Publishers in Court OpenAI and Microsoft have filed competing summary judgment motions in federal court against a coalition of major digital and print news publishers, setting up a definitive legal showdown over whether training generative AI models on copyrighted journalism constitutes fair use. Unsealed evidentiary filings made public in Manhattan federal court reveal unprecedented internal communications from technology executives debating market substitution, licensing structures, and traffic diversion. The legal battle represents a watershed moment for content provenance, digital media economics, and generative AI governance. For publishers, syndication networks, and corporate counsel, the litigation directly challenges the fundamental economic premise of the commercial web: whether tech platforms can ingest original investigative reporting without compensation to generate conversational answers that satisfy user intent directly on third-party interfaces. The dispute highlights a deep structural tension between copyright protections established under the Copyright Act of 1976 and the technical realities of machine learning. Foundational models require vast linguistic corpora containing factual reporting, structured editorial prose, and real-time current affairs to maintain factual grounding and contextual accuracy. As generative answer engines capture search traffic and user attention, the commercial boundary between transformative technological indexing and direct economic market substitution has collapsed into an intense judicial contest. The resulting rulings will establish binding precedents governing data licensing rights, algorithmic fair use, and digital content authenticity worldwide. Fast Facts - **Jurisdiction:** United States District Court for the Southern District of New York (SDNY) - **Lead Defendants:** OpenAI Inc. and Microsoft Corporation - **Plaintiff Coalition:** The New York Times, Daily News, Alden Global Capital publications, and Ziff Davis - **Core Legal Dispute:** Transformative fair use under 17 U.S.C. § 107 vs. market substitution and DMCA § 1202(b) removal of Copyright Management Information (CMI) - **Internal Executive Admissions:** Unsealed memos reveal Microsoft warnings of an AI content "doom loop" threatening essential news suppliers - **Documented Traffic Diversion:** Internal Microsoft data records 83% to 93% drops in click-through rates for major news domains via Copilot answer engines compared to traditional Bing links - **Statutory Damage Exposure:** Billions of dollars in potential statutory liability across millions of cataloged journalistic works ## Judicial Arguments & Evidentiary Deep Dive The competing summary judgment motions unseal years of contentious discovery, exposing the internal friction between rapid AI commercialization and intellectual property rights. According to [MLex](https://www.mlex.com/mlex/articles/2522035/openai-microsoft-clash-with-news-publishers-over-ai-copyright-harm), OpenAI and Microsoft petitioned the court to rule that ingesting public web articles to train transformer weights represents classic transformative fair use, arguing that discovery failed to prove actual demonstrable harm to publisher balance sheets. The defendants rest their legal posture heavily on the landmark precedent of Authors Guild v. Google. In that case, the Second Circuit determined that Google's mass digitization of print books to create a searchable index and display short snippets was a transformative use that did not substitute for the original literary market. OpenAI and Microsoft argue that intermediate copying during pre-training is functionally equivalent: models analyze billions of tokens not to republish expressive prose, but to infer statistical associations, grammatical structures, and factual concepts that comprise human language. However, newly unsealed evidence presented by the news plaintiffs directly attacks the defense of transformative purpose by citing internal admissions from the technology companies' own leadership. Court filings reveal that OpenAI's Head of ChatGPT privately stated that publishers face an "existential threat" from generative products, characterizing consumer AI applications as "largely substitutive, period" and noting that models "will get more and more substitutive as they get better." The plaintiffs argue that these statements directly invoke the Supreme Court's 2023 ruling in Andy Warhol Foundation for the Visual Arts v. Goldsmith, which established that commercial substitution in the primary market of the original work severely undermines a fair use defense. If an AI engine uses original investigative reporting to produce automated answers that eliminate the consumer's need to visit the underlying news domain, the technology operates as an economic substitute rather than a transformative commentary. The evidence is further reinforced by internal Microsoft documentation. In an unsealed strategic assessment, a Microsoft executive cautioned that the company's generative content deployment had initiated a self-destructive dynamic: "Our AI content strategy has started a 'doom loop' that will hurt the performance of our models and the entire web at the same time: It is highly unusual that an end-product threatens the economic foundations of its essential suppliers, but that is the situation we have created for our LLM business with respect to its content supply chain." In addition, Microsoft's internal telemetry revealed that its Copilot answer engine caused an 83% to 93% drop in referral click-through rates for The New York Times and Daily News web properties, alongside a 51% to 94% reduction for Ziff Davis media properties, compared to traditional Bing search results. Under oath, Microsoft Chief Executive Satya Nadella acknowledged that conversational chatbots substitute for traditional browsing by delivering synthesized answers directly on the platform rather than driving users to source publications. Beyond core copyright infringement, publishers are seeking partial summary judgment under Section 1202(b) of the Digital Millennium Copyright Act. The plaintiffs argue that OpenAI intentionally stripped author bylines, copyright symbols, and terms of service during dataset scraping pipelines to prevent automated systems from tracking provenance back to original publishers. ## Comparative Legal Analysis of Fair Use Claims The table below contrasts the opposing statutory arguments evaluated by the Southern District of New York under Section 107 of the Copyright Act: | Statutory Factor | Defense Argument (OpenAI & Microsoft) | Plaintiff Argument (News Publishers) | Judicial & Precedential Tension | |---|---|---|---| | **Factor 1: Purpose & Character** | Highly transformative; statistical learning extracts mathematical weights, not expressive text | Direct commercial substitution; conversational chatbots answer queries to capture ad and subscription revenue | Andy Warhol Foundation v. Goldsmith limits fair use when commercial purposes overlap with original | | **Factor 2: Nature of the Work** | Informational and factual journalism, which receives thinner copyright protection than fiction | Highly creative, expensive investigative reporting requiring significant capital and labor investment | Factual reporting allows broader quotation, but comprehensive wholesale copying weakens the defense | | **Factor 3: Amount & Substantiality** | Copying entire texts is technically necessary to train comprehensive statistical language representations | Ingestion of complete archives without permission, including paywalled and syndicated investigative series | Second Circuit permits full intermediate copying only if necessary for transformative search indexing | | **Factor 4: Market Harm & Substitution** | No evidence of direct market displacement; publishers retain ability to monetize search traffic | Internal data proves 83%–93% referral CTR collapse; destruction of digital licensing and syndication markets | Critical battleground: market substitution represents the central pillar of modern copyright litigation | | **DMCA § 1202(b) CMI Removal** | Incidental loss of metadata during standard web scraping and tokenization pipelines | Willful scrubbing of bylines, copyright notices, and licensing terms to obscure source attribution | Statutory damages range from $2,500 to $25,000 per violation, carrying staggering collective exposure | ## Strategic Takeaways for Media and Technology Executives The unsealed SDNY litigation provides actionable operational guidance for publishing houses, technology vendors, and enterprise compliance teams navigating intellectual property exposure: - **Structure Formal Commercial Licensing Frameworks:** Relying entirely on fair use defenses for commercial AI training carries immense balance-sheet risk. Technology platforms should accelerate bilateral licensing agreements with verified content networks, ensuring sustainable revenue pipelines for high-quality journalism. - **Audit Training Pipelines for Metadata Retention:** Engineering teams compiling pre-training corpora must preserve Copyright Management Information (CMI) and provenance records. Stripping author credits and licensing terms during automated tokenization creates independent statutory exposure under DMCA Section 1202. - **Deploy Precise Content Opt-Out Protocols:** Publishers must implement machine-readable technical barriers beyond basic robots.txt files. Utilizing emerging standards such as WebMCP, structured llms.txt declarations, and authenticated paywall gating provides verifiable evidence of commercial reservation of rights. - **Plan for Hybrid Attribution Architectures:** As courts scrutinize zero-click answer engines, consumer platforms must evolve to include explicit source citations, prominent referral links, and revenue-sharing mechanisms for syndicated answers. Platforms that fail to provide tangible referral traffic risk aggressive regulatory and legal intervention. The legal clash between OpenAI, Microsoft, and the publishing industry demonstrates that the era of unconstrained data scraping has come to an end. Organizations that establish sustainable licensing structures and maintain transparent provenance will build resilient AI systems that respect intellectual property while driving continued technological progress.