On September 4, 2026, two privately held newspaper groups — The Seattle Times, based in Seattle, and Newsday, based on Long Island, New York — jointly filed a 38-page complaint in federal court in Manhattan, suing OpenAI and Microsoft. The complaint accuses the two companies of unlawfully scraping tens of thousands of news articles, including content behind paywalls, to train ChatGPT, Microsoft Copilot, and Bing AI features. The plaintiffs are asking the court to order the destruction of all training datasets and AI models containing their content, and are seeking unspecified damages.
This is the latest round of copyright pushback by the media industry against AI companies in recent years, and it is also the first case to date to bring trademark infringement claims alongside copyright claims. The plaintiffs allege that OpenAI’s models can generate fabricated content falsely attributed to The Seattle Times and Newsday, damaging the brand credibility of both media outlets.
Not Just Copyright: False AI Attribution Is a New Threat
In previous lawsuits of this kind, media plaintiffs have typically focused on the traditional copyright framework of “content being copied.” What is new in this complaint is that it explicitly accuses AI outputs of causing “confusingly misleading” harm to media brands. When a user asks ChatGPT a question and receives a passage that appears to be quoted from The Seattle Times but is in fact entirely fabricated, the damage to the outlet’s credibility is real, yet difficult to quantify through declines in clicks or advertising revenue. This opens a new dimension for future litigation strategies.
The complaint cites one industry data point: In December 2025, search referral traffic to midsized media outlets fell 47% year over year. This figure is not rhetorical exaggeration by the legal team — according to Chartbeat data, search traffic to small publishers fell by more than 60% over the same period, while referrals from Google to media websites shrank by about 33% worldwide. AI-generated summary responses are systematically severing the click that once connected users to original reporting.
“Like a snake devouring its own tail — generative AI trained on carefully produced content that requires substantial resources is now directly competing with news organizations through AI-generated substitute content, threatening to destroy those very news organizations.” — from the complaint
The Timing of the Lawsuit Is Notable
The decision to file the complaint on September 4, 2026, came at a sensitive moment. Just one day before the lawsuit was filed, the U.S. Department of Justice submitted a court brief in support of OpenAI in The New York Times v. OpenAI, explicitly arguing that training AI systems on news content constitutes fair use. This was the first time the federal government had publicly intervened in this type of case, and it did so on the side of the AI company — an important judicial signal for all media organizations pursuing litigation.
In other words, The Seattle Times and Newsday chose to sue despite knowing that the government’s stance was unfavorable. The logic behind this is not only legal maneuvering, but also likely a proactive move in the battle for public discourse: at a critical moment when OpenAI is pushing toward an IPO at a trillion-dollar valuation, the lawsuit forces the question of “who pays for journalism” into public view.
Microsoft spokesperson Drew Born said in a statement: “We are surprised by this lawsuit, but we understand the importance of local journalism and are always willing to sit down and discuss solutions.” OpenAI responded that its models are trained on publicly available data and comply with fair use principles, but did not comment further on the specific lawsuit.
Litigation or Negotiation: A Fork in the Road for the Media Industry
The media industry is currently split between two sharply different strategies. The New York Times, the Chicago Tribune, The Denver Post, and others have chosen the litigation path; The Washington Post and News Corp — whose holdings include The Wall Street Journal and the New York Post — have chosen to sign licensing agreements with OpenAI. This divide itself is a signal: there is no unified “media position,” only each company’s calculation of its own business circumstances.
Privately held media companies and publicly listed groups have structural differences here. The Seattle Times and Newsday are both privately held and face no pressure from quarterly reporting, allowing them to withstand a longer litigation cycle. For publicly listed media groups, negotiating a licensing fee with OpenAI can immediately improve financial results, while any damages from litigation may not arrive until years later.
Destroying Training Data: Impossible Remedy or Bargaining Chip?
The complaint asks the court to order the “destruction and/or impoundment” of datasets and AI models containing the plaintiffs’ content. In legal reality, this demand is almost impossible to enforce in full — OpenAI’s training data and model weights have long since been blended together and cannot be cleanly separated like deleting a folder. But that does not mean the demand has no value: it is a high starting point at the negotiating table. Even if the court does not support full destruction, it may still order limits on the commercial use of certain content or require the establishment of licensing and revenue-sharing mechanisms.
The New York Times’ 2023 lawsuit is still ongoing. The judge has already declined to grant OpenAI’s preliminary motion to dismiss, and the case is entering the discovery phase — meaning OpenAI must disclose the composition of its training data to the court. This is consequential for the entire industry, not just for The New York Times.
The Real Question: Who Bears the Cost of Producing Journalism?
Alan Fisco, president and CEO of The Seattle Times, wrote in an internal email to employees that the company spends millions of dollars each year producing news content and “must defend our content from being used without compensation.” He emphasized that the lawsuit is “not intended to hinder AI innovation, but to ensure that innovation does not come at the expense of our newspaper’s business model.”
That statement touches the core contradiction of this dispute: large AI companies use news content to train models, and the trained models then replace users’ visits to news websites with summary-style question-and-answer responses. This closed loop is commercially coherent, but ethically parasitic. OpenAI is valued at about $850 billion and is preparing to go public at a trillion-dollar valuation, while search referral traffic to midsized media outlets has already fallen by nearly half.
The fair use doctrine itself has a long history, and its boundaries have always evolved dynamically under the impact of new technologies. Google’s book-scanning project, which digitized millions of books for search purposes, was ultimately ruled by courts to be fair use. Whether AI training follows the same logic still has no Supreme Court-level precedent. The Justice Department’s support for OpenAI reflects the executive branch’s inclination, but it is not binding as a final judicial ruling.
Assessment
This case is highly likely to end in an out-of-court settlement, because both sides have strong incentives to avoid a final judgment — OpenAI does not want a court to find its training data unlawful before its IPO, while the media plaintiffs need stable cash flow rather than remote damages years down the road. The real battlefield remains The New York Times case: once that lawsuit enters discovery, the industry will for the first time obtain a judicially tested “AI training data audit report.” At that point, which content was scraped, how often it was scraped, and whether technical restrictions were bypassed will no longer be matters of speculation, but part of the legal record.
The lawsuit by The Seattle Times and Newsday is more like an act of strategic pressure: securing bargaining chips before an industry-defining ruling arrives. The content ledger between media outlets and AI companies will have to be settled sooner or later; only the method remains undecided.
© 2026 Winzheng.com 赢政天下 | 转载请注明来源并附原文链接