Nearly 400 local newspapers sue OpenAI, Microsoft over alleged copyright theft

Nearly 400 Local Newspapers Sue OpenAI and Microsoft Over Alleged Copyright Theft — What It Means for AI's Legal Future

On June 24, 2026, a coalition of publishers collectively operating nearly 400 local and regional newspapers filed a federal copyright lawsuit against OpenAI and Microsoft. Their allegation: the companies systematically scraped their articles without permission and used them to train ChatGPT and Copilot. This is the 26th lawsuit against OpenAI and the 11th against Microsoft in an accelerating wave of AI copyright litigation. For the first time, a community news coalition has mobilized at this scale, and they've introduced allegations about paywall scraping and copyright metadata stripping that go beyond the factual record established in the prominent New York Times complaint. The outcome could reshape how AI companies are permitted to ingest web content at scale.

Who Filed, Where, and When: The Coalition Behind the Lawsuit

The complaint is titled Richner Communications, Inc. v. Microsoft Corp. and was filed in the U.S. District Court for the Southern District of New York (Case No. 1:26-cv-05320) on June 24, 2026. Richner Communications, Inc. — a Long Island-based publisher — leads the charge, joined by dozens of other local and regional publishers including AIM Media Indiana Operating, LLC, AIM Media Midwest Operating, LLC, AIM Media Texas Operating, LLC, The New York Amsterdam News, Arkansas Democrat-Gazette, Inc., The Ogden Newspapers, Inc., and Concord Publishing House, Inc., among others. Together, these publishers own and operate nearly 400 newspapers spread across communities in multiple states. It is one of the broadest multi-publisher actions the litigation wave has seen so far.

Platkin LLP is representing the publishers — a mission-driven firm founded by former New Jersey Attorney General Matthew J. Platkin, who served as the state's 62nd AG from February 2022 to January 2026. The firm's founding team draws on senior litigators from the New Jersey Attorney General's Office, including Angela Cai — who served most recently as Executive Assistant Attorney General (the third-ranking position in the office) and previously as Deputy Solicitor General, where she helped build New Jersey's appellate practice — and Ravi Ramanathan, the inaugural Director of New Jersey's SAFE (Statewide Affirmative Firearms Enforcement) Office. The publishers chose the Southern District of New York for good reason: it's where much of the AI copyright litigation is already happening, including the high-profile New York Times Co. v. OpenAI action filed in December 2023 and other consolidated cases.

Legal observers predict this suit will most likely be stayed while the courts handle summary judgment motions in the ongoing multidistrict litigation (MDL) known as In re: OpenAI, Inc. Copyright Infringement Litigation. That MDL — consolidated by the Judicial Panel on Multidistrict Litigation in April 2025 and assigned to Judge Sidney H. Stein in the U.S. District Court for the Southern District of New York (Case No. 1:25-md-03143) — is working through threshold legal questions about fair use, transformative use, and what training data practices should be permissible. If those bigger questions get resolved at the MDL level, it could potentially resolve multiple suits at once.

The Core Allegations: Scraping, Paywall Violations, and DMCA Violations

The complaint builds its case on three causes of action: direct copyright infringement, vicarious copyright infringement, and violation of the DMCA's copyright management information provisions. Woven through the infringement claims is a striking factual narrative — that defendants scraped content from behind paywalls, not just freely available web pages. Together, these theories paint a picture of what the publishers describe as systematic exploitation of local journalism designed to evade detection and licensing negotiations.

1. Direct and Vicarious Copyright Infringement

The publishers allege that OpenAI and Microsoft systematically and secretly crawled hundreds of their news websites, copying vast quantities of original reporting, investigative articles, breaking news, and editorial content to build the training datasets for ChatGPT and Microsoft Copilot. This wasn't incidental web crawling, they argue — it was deliberate harvesting of news content at a massive scale. Potentially millions or billions of articles were copied without any licensing agreement, compensation, or even notification.

The complaint also alleges vicarious infringement — a distinct legal theory holding that a party can be liable for infringement it did not directly commit if it had the practical ability to supervise or limit the infringing activity and derived a direct financial benefit from it. In practice, this means the publishers are arguing that even if certain scraping was carried out by contractors or automated systems, OpenAI and Microsoft retained sufficient control over those systems and reaped sufficient commercial reward from the resulting models to be held responsible as if they had done the copying themselves. Vicarious liability doctrine, established in cases like Fonovisa, Inc. v. Cherry Auction, Inc., 76 F.3d 259 (9th Cir. 1996), and later affirmed in the Supreme Court's MGM Studios, Inc. v. Grokster, Ltd., 545 U.S. 913 (2005), is well-established in copyright law, making this a credible secondary theory even if the direct infringement claim faces challenges.

2. Paywall Scraping: A Factual Allegation That Undermines the "Publicly Available" Defense

Here's where this case's factual record stands apart. The publishers have invested heavily in building and maintaining paywalls, metered access models, registration walls, and other tools to protect and monetize their original reporting. According to the complaint, OpenAI and Microsoft didn't just crawl free homepage content. As the publishers' filing puts it, they "spent billions of dollars to protect the work — including by shielding it behind paywalls — all for naught, as the defendants took all of it." Content that was never offered for free web indexing was allegedly copied anyway.

This allegation strikes at the heart of OpenAI's primary defense. The company has consistently argued it trained on "publicly available data" accessed through ordinary web scraping. But if it scraped content that was gated behind access controls? That story falls apart. While paywall circumvention is not itself a separate count in the complaint, it is a central factual allegation that directly undermines OpenAI's fair use framing — potentially the most damaging factual claim in the entire filing if it survives discovery.

3. DMCA Copyright Management Information (CMI) Violations

The lawsuit also invokes the Digital Millennium Copyright Act (DMCA), specifically its provisions on copyright management information codified in 17 U.S.C. § 1202. The publishers allege that OpenAI knowingly stripped or removed CMI from their works — including author bylines, reporter names, publication dates, and copyright notices — as part of its data preprocessing routines. Under the statute, intentionally removing CMI with knowledge that it facilitates or conceals infringement carries its own statutory penalties, separate from standard copyright damages. This matters more than it might initially seem. A CMI violation can result in substantial statutory damages per violation — the statute sets a range that can reach up to $25,000 per violation in willful cases. Those numbers can dwarf what most fair use analyses would suggest.

Why the CMI claim matters for developers and the industry: If courts rule that stripping author metadata, publication dates, and copyright notices from scraped text during AI training constitutes a DMCA violation, it will force a fundamental redesign of data preprocessing pipelines. Any organization — from a major AI lab to a smaller company fine-tuning an open-source model on Common Crawl data — could face statutory liability for CMI removal that happens as a routine byproduct of text normalization, JSON parsing, or deduplication. Suddenly, it becomes legally necessary to preserve and transmit copyright metadata through every stage of model training. That raises engineering complexity and creates potential liability exposure for vendors of training datasets.

The Local Journalism Angle: Why This Coalition Represents a Fundamentally Different Case

Nearly 400 local newspapers suing together isn't just an impressive headline. It reflects a structural vulnerability in community journalism that makes this case fundamentally different from, and arguably more consequential than, the New York Times lawsuit. The Times is a global media brand with substantial legal resources, a diversified revenue base spanning digital subscriptions, merchandise, events, and advertising. It can absorb legal costs and disruptions. The publishers in this coalition operate on razor-thin margins in shrinking local advertising markets. Many have single-digit newsroom staffs and revenue models that depend almost entirely on local advertising and, increasingly, small-dollar digital subscriptions.

These publishers have invested heavily in original reporting — the kind of granular, place-specific coverage of city council meetings, school board decisions, local business developments, and community events that can't be replicated by a wire service or synthesized from national headlines. That reporting is precisely what makes local newspaper archives valuable training data for AI models seeking broad geographic and topical coverage. Yet, according to the complaint, not a cent of the value extracted has flowed back to these publishers. No licensing negotiations were ever proposed.

The human cost is not abstract. Local newsrooms have been closing at an accelerating rate for more than a decade, driven by the migration of classified advertising to Craigslist and digital platforms, display advertising fragmenting to Google and Facebook, and declining print subscriptions. Research tracking these trends shows that hundreds of U.S. counties are now news deserts — communities with no dedicated local news coverage at all. Residents rely on social media rumors or regional outlets that lack deep knowledge of local affairs. Publishers argue that AI companies are committing a double injury: first, unpaid extraction of their content; second, substitution of that content with AI-generated alternatives that displace visits and subscription conversions. Instead of driving traffic to newspaper websites, AI chatbots surface summaries and excerpts that serve users without them ever needing to click through.

The publishers' filing frames the stakes in stark terms, describing the unchecked AI content scraping as "a death knell for local journalism — which remains the most trusted news source in America." That framing gives this litigation a public-interest dimension that the New York Times case, brought by a well-resourced global brand, could not claim in the same way.

OpenAI and Microsoft's Defense Position and Fair Use Arguments

OpenAI spokesperson Drew Pusateri responded to the lawsuit with what has become the company's standard legal framing, telling Bloomberg Law: "Our models empower innovation, are trained on publicly available data, and are grounded in fair use." Microsoft did not immediately respond to a request for comment, per Bloomberg Law's reporting on the filing.

The fair use defense has been OpenAI's primary legal shield across all copyright litigation involving training data. It rests on four statutory factors codified in 17 U.S.C. § 107: (1) the purpose and character of the use; (2) the nature of the copyrighted work; (3) the amount and substantiality of the portion used; and (4) the effect of the use on the potential market for or value of the original work. OpenAI's position is that training an LLM is a transformative use — analogous to how a search engine indexes content or how researchers might use copyrighted materials for analysis. Therefore, it argues, LLM training does not substitute for or cannibalize the original works in the market. The publishers vigorously dispute this. They argue that AI-generated summaries and responses directly cannibalize their reader traffic and subscription revenue by providing users with information extracted from their articles without requiring them to visit the newspaper's website or subscribe.

The paywall scraping allegation, if it survives discovery, may prove to be the most legally challenging aspect of OpenAI's fair use defense. Historically, courts have been more sympathetic to fair use claims involving freely available public content than to claims involving access that bypasses technical protection measures or access controls. The DMCA imposes stricter liability standards for circumvention and makes clear that Congress intended to protect technological access controls even beyond the scope of copyright itself. This creates potential two-front legal exposure: copyright infringement liability for the unauthorized copying, plus separate DMCA liability for stripping copyright management information. Even if a court were inclined to find fair use on the copyright claim, the CMI violations could stick independently.

Where This Lawsuit Sits in the Broader AI Copyright Litigation Landscape

This is the 26th lawsuit against OpenAI and the 11th against Microsoft over AI training data practices and alleged copyright infringement. The sheer volume and complexity of overlapping litigation has prompted the federal judiciary to begin consolidating cases under the multidistrict litigation (MDL) mechanism. The primary MDL, In re: OpenAI, Inc. Copyright Infringement Litigation (Case No. 1:25-md-03143, S.D.N.Y.), was formed when the Judicial Panel on Multidistrict Litigation consolidated twelve overlapping actions in April 2025 and assigned them to Judge Sidney H. Stein in the Southern District of New York. The Richner case is expected to be folded into that process or stayed pending the MDL's outcome.

Here's a comprehensive look at how the major publisher-led and creator-led AI copyright actions compare:

Case Name	Plaintiffs	Filed	Court / Jurisdiction	Key Legal Claims	Status (as of late June 2026)
New York Times Co. v. OpenAI, Inc. & Microsoft Corp.	The New York Times Company (single plaintiff)	December 2023	U.S. District Court, S.D.N.Y.	Copyright infringement, DMCA CMI violations, unjust enrichment	Active; proceedings ongoing
In re: OpenAI, Inc. Copyright Infringement Litigation (MDL No. 3143)	Multiple authors, individual creators, small publishers (class action framework)	Consolidated April 2025	U.S. District Court, S.D.N.Y. (Judge Sidney H. Stein; Case No. 1:25-md-03143)	Copyright infringement, class-based claims, potential damages class certification	Summary judgment briefing; considered bellwether for fair use questions
Richner Communications, Inc. v. Microsoft Corp. & OpenAI, Inc.	Dozens of publishers collectively operating nearly 400 newspapers	June 24, 2026	U.S. District Court, S.D.N.Y. (No. 1:26-cv-05320)	Direct copyright infringement, vicarious copyright infringement, DMCA CMI removal	Newly filed; likely stay pending MDL summary judgment ruling

The proliferation of suits is itself a deliberate litigation strategy. Each new plaintiff group adds distinct factual circumstances — different publication types, different geographic markets, different alleged instances of verbatim reproduction or paywall bypass, different industries (news vs. books vs. software code). This diversity makes it harder for OpenAI and Microsoft to secure a single, sweeping fair use ruling that might end all litigation at once. Courts are more inclined to narrow rulings when facts vary significantly across cases, meaning the defendants cannot simply prevail once and have that victory apply universally.

Practical Implications for Developers, AI Teams, and the AI Ecosystem

For engineers and product teams building applications on top of OpenAI or Microsoft APIs, for independent researchers training their own large language models, and for companies developing foundation models, the legal landscape being shaped by suits like this one has direct and material practical implications:

• Training data provenance documentation will become non-negotiable. If the CMI stripping claims succeed in court, any data pipeline that processes web-scraped text and discards metadata as a preprocessing step could be legally implicated. Responsible AI development will increasingly require retaining copyright management information and source metadata through the entire training process — not just for legal cover but for model transparency, auditability, and compliance. Teams will need to modify their standard preprocessing routines to preserve this information, which will increase storage and computational complexity.
• Paywalled and access-controlled content is emphatically not fair game for training. The explicit allegation that OpenAI and Microsoft crawled behind paywalls will force any serious AI team to audit how their web-scraping infrastructure handles access-controlled and restricted pages. Crawling paywalled content — whether through cookie manipulation, exploitation of cached copies, or other technical means — is now clearly in legal and regulatory crosshairs. Any organization planning to build training datasets from web data must implement explicit technical controls to respect robots.txt, honor HTTP 401/403 authentication responses, and refrain from circumventing access controls.
• Publisher licensing deals will accelerate dramatically, raising costs and limiting access. OpenAI has already struck content licensing agreements with major publishers including The Associated Press, Axel Springer, and Condé Nast. Suits like this one increase both the legal risk and the reputational cost of not having a licensing deal in place for any major publication. Expect the pace of such licensing negotiations to accelerate sharply, and expect smaller regional publishers and local news organizations to begin organizing collectively to negotiate group licensing terms. This will increase the cost of training data for AI companies and potentially create new barriers to entry for smaller competitors who cannot afford the licensing fees.
• Fair use doctrine is not settled law for LLM training — treat assumptions with skepticism. Developers who have been operating under the assumption that training on broadly scraped web data is unambiguously protected by fair use should recognize that assumption as deeply contested by courts, not yet confirmed by definitive appellate precedent. Courts have not yet issued a final, authoritative ruling on whether LLM training qualifies as fair use under all or most circumstances. Until they do, the legal risk remains real, material, and quantifiable. Companies should plan for scenarios in which courts find against fair use, or carve out narrow exceptions that limit its scope.
• The MDL summary judgment ruling is the linchpin of the entire legal landscape. Because this case will very likely be stayed or consolidated with the MDL pending its summary judgment phase, the MDL's outcome is arguably the single most important near-term legal event for the entire AI industry. A federal judge's ruling that LLM training does not qualify as fair use — or that it does but with significant restrictions — would send immediate shockwaves through every company that has built production models on large web corpora. Conversely, a ruling strongly affirming fair use could temporarily insulate the industry from the most aggressive copyright claims, though targeted statutes or legislation could still follow.

Key Takeaways and Legal Summary

• Scale is unprecedented for local publishers: Dozens of publishers representing nearly 400 local and regional newspapers filed suit on June 24, 2026, in the U.S. District Court for the Southern District of New York (Case No. 1:26-cv-05320) — the largest and most comprehensive coordinated legal action by community news organizations against AI companies to date.
• Three-count legal attack targets different vulnerabilities: The complaint alleges direct copyright infringement and vicarious copyright infringement for unauthorized copying at scale, plus DMCA violations for stripping author bylines and copyright metadata from scraped articles. Woven through the infringement counts is the factual allegation that defendants scraped content from behind paywalls — directly undercutting the "publicly available data" defense.
• CMI (Copyright Management Information) removal is a sleeper legal issue with high penalties: The allegation that OpenAI knowingly stripped copyright management information — including author bylines, publication dates, and copyright notices — adds a legally distinct DMCA claim under 17 U.S.C. § 1202 with substantial statutory penalties that can reach up to $25,000 per violation in willful cases, sitting on top of standard infringement damages and potentially providing significant leverage in settlement discussions.
• OpenAI's standard fair use defense faces its toughest factual test: Spokesperson Drew Pusateri's invocation of "publicly available data" and "fair use" carries substantially less weight when the complaint alleges the defendants scraped behind paywalls, directly contradicting the "publicly available" premise of that defense.
• This is the 26th lawsuit against OpenAI, the 11th against Microsoft: The sheer volume and diversity of litigation signals that the industry cannot wait for a single, decisive court ruling. Teams and organizations should plan for a multi-year period of legal uncertainty, regulatory flux, and cost escalation.
• A stay of this particular case is probable: The Richner case will probably be paused or consolidated pending summary judgment and other key rulings in the MDL (In re: OpenAI, Inc. Copyright Infringement Litigation, No. 1:25-md-03143, S.D.N.Y.), making that proceeding the critical near-term bellwether for how courts will actually decide the entire AI training data debate.
• Local journalism and community news are the human stakes: Unlike the high-profile New York Times lawsuit, this action foregrounds the existential economic fragility of hundreds of community newsrooms. The publishers' filing describes the AI companies' conduct as a "death knell for local journalism," giving this litigation a political and public-interest dimension that may influence both federal courts and members of Congress, and may resonate with juries in ways that abstract copyright doctrine alone might not.

What Comes Next: Procedural Timeline and Strategic Considerations

The immediate procedural steps will be straightforward. OpenAI and Microsoft will file their responses to the complaint — typically within 21 days of service. They will almost certainly move to dismiss, consolidate with the MDL, or seek an immediate stay. The choice of which motion they file first will itself signal strategic intent. A motion to dismiss signals confidence in the legal theories, while a move to consolidate or stay signals a desire to avoid litigating facts in a potentially unfavorable forum.

The more consequential development to watch closely is the MDL's trajectory toward summary judgment. At that stage, a federal judge will have to engage directly and substantively with the fair use question. Summary judgment briefing and oral arguments will be the moment when the full legal and policy arguments on both sides get their most thorough judicial review. Meanwhile, legislative pressure is building in parallel. Multiple members of Congress have floated AI training data transparency bills, AI safety bills with training data disclosure requirements, and bills that would limit or prohibit scraping of paywalled content for training purposes. A judicial ruling against OpenAI — or even a partial ruling in favor of plaintiffs on the DMCA CMI issue alone — could supercharge and accelerate that legislative effort, creating a pincer movement of judicial and statutory restriction.

For the nearly 400 local newspapers that filed suit, the outcome is existential in the most literal sense. If they prevail — whether on the full copyright claim, the paywall-scraping factual theory, or the DMCA violation — it establishes that community journalism has a compensable legal right to its archives in the AI age. It opens the door to damages that could, at least partially, offset the revenue lost to AI competition and the costs of AI litigation defense. If they lose, the precedent would effectively green-light AI companies to continue training on local news content indefinitely, at scale, without payment or licensing. For many of these publications already operating on razor-thin margins, that result could prove as damaging as the advertiser exodus and circulation collapse of the past decade combined.