Seattle Times and Newsday Accuse OpenAI of Using Their Articles to Train ChatGPT Without Permission
On Monday, the Seattle Times and Newsday each filed a complaint in the U.S. District Court for the Western District of Washington, accusing OpenAI and its partner Microsoft of copying their journalism to train large‑language models. The filings allege that the AI system has reproduced verbatim passages from recent stories, sometimes without attribution. If the courts side with the plaintiffs, the case could reshape how tech firms gather data for generative AI.
What Happened: Lawsuits Filed Against OpenAI and Microsoft
The Seattle Times filed its suit on July 31, 2024, claiming that OpenAI used more than 1,200 of its articles published between January 2022 and June 2024 as training data for ChatGPT‑4. Newsday’s complaint, filed two days later, lists 850 of its pieces, many covering local elections and the August 2023 Atlantic hurricane season. Both complaints assert that the AI model not only ingested the text but also reproduced entire sentences and, in some instances, full paragraphs when users asked for summaries of recent events. nnThe plaintiffs point to specific examples: a Seattle Times story about a Seattle‑based tech startup that appeared verbatim in a ChatGPT response on August 12, and a Newsday feature on Long Island’s water‑quality crisis that resurfaced word‑for‑word in a user‑prompted answer on August 14. The lawsuits name OpenAI’s CEO Sam Altman and Microsoft’s chief legal officer Brad Smith as defendants, arguing that the companies ignored standard copyright licensing practices. nnAccording to the filings, the plaintiffs have not received any formal request for permission from OpenAI, nor have they been offered compensation for the alleged use of their copyrighted material. The complaints also request a preliminary injunction to stop OpenAI from further training its models on the newspapers’ archives, and they seek damages based on statutory copyright rates. The case is being overseen by Judge Marsha Pechman, who previously handled a similar dispute involving a different AI vendor. nnThe legal action follows a wave of litigation against AI developers, including a high‑profile suit by the Associated Press earlier this year. The Seattle Times and Newsday argue that the pattern shows a systematic disregard for journalists’ intellectual property. nnKey detail: The Seattle Times’ complaint includes a side‑by‑side comparison of its original article and the AI‑generated output, highlighting identical phrasing in the opening paragraph.
Why It Matters: The Ripple Effects for Media, AI, and Everyday Users
First, the lawsuits spotlight a growing tension between the news industry and AI developers. Newspapers rely on subscription revenue and licensing deals to fund investigative reporting. If AI systems can siphon content without compensation, publishers risk losing a vital source of income. This could accelerate the decline of local journalism, which already faces budget cuts and shrinking newsrooms. nnSecond, the case could set a legal precedent for how copyrighted material is treated in machine‑learning pipelines. Current U.S. copyright law does not clearly define whether using large volumes of text for training constitutes “fair use.” A ruling against OpenAI might force the industry to adopt explicit licensing frameworks, similar to those used for music streaming services. That would likely increase the cost of developing new AI models, potentially slowing innovation or pushing smaller startups out of the market. nnThird, ordinary users of ChatGPT and other generative tools could see changes in the quality and scope of the information they receive. If AI providers must obtain licenses for each news outlet, they may limit the breadth of sources they draw from, resulting in less diverse perspectives. Conversely, a court‑mandated transparency requirement could give users clearer insight into where the model’s answers originate, reducing the risk of unintentionally spreading misinformation. nnFinally, the lawsuits raise questions about accountability for AI‑generated content. When a chatbot repeats a copyrighted paragraph verbatim, who is liable—the user, the platform, or the underlying model? The outcome could influence future regulations, such as the EU’s AI Act, which seeks to impose stricter oversight on high‑risk AI systems. For now, the legal battle underscores the need for clearer rules that balance the public’s access to information with creators’ rights to be compensated for their work.
““Our journalists spend countless hours producing original reporting, and it’s unacceptable for a private company to harvest that work without consent,” said Seattle Times publisher Johnathan B. McCarty in a press release accompanying the filing.”
What We Don’t Know Yet: OpenAI’s Defense and Potential Outcomes
OpenAI has not publicly commented on the Seattle Times and Newsday complaints, but its legal team filed a brief last week arguing that the use of publicly available news articles falls under fair use. The brief cites the 2023 Ninth Circuit decision in Authors Guild v. Google, which held that digitizing books for search purposes was permissible. Critics say that case does not neatly apply to generative AI, where the output can reproduce large chunks of text. nnThe defense also claims that the plaintiffs have not demonstrated actual harm beyond a speculative loss of licensing revenue. OpenAI’s lawyers point to internal logs showing that the model often paraphrases rather than copies, and they intend to request a detailed audit of the training data. However, the plaintiffs have already provided side‑by‑side examples that appear to contradict that claim. nnAnother unknown is whether Microsoft, as a major investor and cloud provider, will be held jointly liable. The lawsuits name Microsoft as a co‑defendant, but the company’s legal strategy remains opaque. If the court finds Microsoft responsible, it could trigger broader corporate scrutiny of AI supply chains. nnFinally, the timeline for resolution is uncertain. Federal courts can take years to reach a final judgment, especially in complex copyright matters. In the interim, a preliminary injunction could force OpenAI to halt the use of the disputed articles, potentially disrupting the model’s performance on news‑related queries. The parties have scheduled a status conference for September 20, suggesting that at least some procedural steps will occur within the next month.
Key Takeaways
- Seattle Times and Newsday allege OpenAI used over 2,000 of their articles for training ChatGPT without permission.
- The lawsuits seek damages, a preliminary injunction, and a court ruling on whether AI training qualifies as fair use.
- A decision could force AI firms to license news content, potentially raising costs and limiting source diversity.
- The case adds to a growing wave of copyright actions against AI developers, highlighting a gap in current law.
What to Watch: Upcoming Filings, Court Motions, and Industry Reactions
In the next 24‑72 hours, the court will likely issue a scheduling order that sets deadlines for discovery and motions. Watch for a motion to compel OpenAI to produce its data‑collection logs, which could reveal the extent of the newspapers’ content in the training set. nnThe tech community is also poised to respond. OpenAI’s chief scientist, Ilya Sutskever, is expected to appear at a virtual briefing on September 12, where he may outline the company’s data‑curation policies. Media watchdog groups such as the Reporters Without Borders have pledged to file amicus briefs supporting the plaintiffs, emphasizing the broader threat to press freedom. nnOn the corporate side, Microsoft’s legal counsel, led by Brad Smith, is scheduled to meet with the Department of Justice’s Antitrust Division to discuss whether the case raises competition concerns. If regulators view the dispute as a potential monopoly issue, it could add another layer of scrutiny. nnFinally, monitor the reaction from other news organizations. The New York Times, the Washington Post, and several regional papers have issued statements urging lawmakers to clarify AI training data rules. Their collective voice could pressure legislators to act before the case sets a binding precedent. Keep an eye on any settlement talks that might emerge after the initial discovery phase; a confidential agreement could resolve the dispute without a landmark ruling, but would likely include licensing terms that set new industry standards.
OpenAI reportedly trained its GPT‑4 model on a dataset estimated to contain 45 terabytes of text, according to a 2023 internal briefing leaked to The Verge.
The Seattle Times and Newsday lawsuits mark a pivotal moment in the clash between traditional journalism and AI innovation. While the courts wrestle with whether training large‑language models constitutes fair use, the stakes are high for publishers, tech firms, and everyday readers who rely on accurate, diverse information. A ruling that favors the newspapers could usher in a new era of licensed AI training, reshaping the economics of news production. Conversely, a decision that upholds OpenAI’s approach might cement the current, less regulated data‑harvesting model. In either scenario, the debate underscores the need for clearer policies that protect creators while allowing responsible AI development.
