The debate over fair use in artificial intelligence training has reached a boiling point, with Microsoft and OpenAI doubling down on their legal stance even as critics warn that their products are sounding the “death knell for local journalism.” The two companies have repeatedly argued that using copyrighted news articles, books, and other published works to train large language models like ChatGPT and the underlying GPT-4 system constitutes fair use under U.S. copyright law. This position has drawn fierce opposition from news organizations, authors, and advocacy groups who claim it enables mass appropriation of intellectual property without compensation or credit.
The Fair Use Defense in AI
Fair use is a legal doctrine that allows limited use of copyrighted material without permission for purposes such as criticism, comment, news reporting, teaching, scholarship, or research. The determination hinges on four factors: the purpose and character of the use, the nature of the copyrighted work, the amount and substantiality of the portion used, and the effect on the potential market. Microsoft and OpenAI argue that training AI models on publicly available text is transformative—the AI does not reproduce the original work but learns patterns to generate new content. They also contend that the training data is not directly disclosed to end users, so no market harm occurs.
However, news publishers see a different reality. When a user asks ChatGPT to summarize a recent news article or generate a story in the style of a specific journalist, the system often produces outputs that closely paraphrase or reproduce key details from copyrighted articles. In some cases, chatbots have been caught reproducing entire paragraphs verbatim. This has led to multiple high-profile lawsuits, including a consolidated case brought by The New York Times, which alleges that OpenAI and Microsoft “unlawfully copied millions of articles” to build their AI products. The lawsuit argues that the companies could have struck licensing agreements but chose to bypass them, thereby threatening the economic viability of journalism.
Impact on Local Journalism
The phrase “death knell for local journalism” is not hyperbole. Local newspapers have been under severe financial pressure for decades due to the decline of print advertising and the rise of digital platforms like Google and Facebook, which capture the vast majority of digital ad revenue. Now, AI-powered chatbots are exacerbating the problem by delivering news summaries directly to users without sending them to publisher websites. This—combined with AI-generated “news” websites that produce low-quality, algorithmically assembled articles—reduces traffic to legitimate local outlets, further shrinking their already thin revenue streams.
According to a report from the Northwestern University Medill School of Journalism, the United States has lost nearly 2,900 newspapers since 2005, and many of the remaining are “ghost newspapers” with severely reduced staff. The rise of generative AI threatens to accelerate this trend. If readers can get news from a chatbot without clicking on a publisher’s page, the financial model that supports fact-checking, investigative reporting, and community coverage collapses. Local journalists are already seeing the effects: some newsrooms have reported that their articles are being scraped and repackaged by AI chatbots within minutes of publication, leaving the original outlet with no traffic or attribution.
Legal and Regulatory Responses
Several countries are moving to regulate AI training data. The European Union’s AI Act includes transparency requirements for general-purpose AI models, mandating that developers disclose any use of copyrighted materials. In the United States, the Copyright Office has initiated a study on AI and copyright, and lawmakers have introduced bills such as the “AI Foundation Model Transparency Act” and the “Copyright Office Artificial Intelligence Act.” Meanwhile, publishers are increasingly turning to licensing agreements. News Corp, Axel Springer, and Le Monde have signed multiyear, multi-million-dollar deals with OpenAI that allow the company to use their content for training and display. However, these agreements cover only a fraction of the news ecosystem, leaving thousands of smaller local outlets without any compensation.
Microsoft and OpenAI argue that requiring licenses for all training data would stifle innovation and favor large incumbents, while making it impossible for smaller AI startups to compete. Yet critics counter that the current approach amounts to “tech colonization” of creative industries, where the tech giants extract value from the work of journalists, writers, and artists without sharing the proceeds. The recent finding by Nobel Prize-winning economist Sir Angus Deaton and others that AI could worsen inequality has added weight to the argument that the economic benefits of AI must be distributed more broadly.
Ethical and Practical Considerations
Beyond the legal dispute, ethical questions loom. Many journalists and scholars argue that using copyrighted news articles to train commercial AI products without consent violates fundamental norms of fairness and attribution. Even if the practice is technically legal under a broad reading of fair use, they say it undermines the social contract between creators and the platforms that profit from their work. The debate also touches on the quality of AI-generated news: because large language models are trained on the full corpus of internet text, they can reflect biases, propagate misinformation, and fabricate details—hallucinations that are particularly dangerous in news contexts where accuracy is paramount.
Local journalism is uniquely vulnerable because it relies on trust and geographic relevance. A chatbot cannot attend a city council meeting or interview a small business owner. It can only synthesize what already exists online, often missing nuance and local context. If the financial foundation of local news erodes further, the ability to hold local institutions accountable—school boards, town councils, police departments—diminishes. This in turn can lead to decreased civic engagement and weakened democracy at the grassroots level.
Industry Reactions and the Path Forward
Some news executives have called for a collective licensing regime similar to the one that exists for recorded music, where streaming services pay royalties to artists through organizations like ASCAP and BMI. The News Media Alliance, which represents over 2,000 publications, has been lobbying for legislation that would allow publishers to negotiate collectively with tech platforms. In a recent statement, the Alliance said that “AI companies must obtain permission and pay fair value for the use of news content.”
OpenAI has responded by creating a “Media Manager” tool that would allow publishers to specify how their content can be used in AI training, but the tool has not yet been fully deployed. Microsoft has launched initiatives to support local news, including grants and the AI-assisted newsroom program, though critics argue these efforts are small relative to the harm inflicted by the widespread scraping of content. The tension between the two companies’ fair use position and their sustainability rhetoric is growing increasingly untenable.
As the legal battles play out, the future of local journalism hangs in the balance. The next few years will determine whether the fair use doctrine is stretched to cover AI training at the expense of original content creators, or whether new legal frameworks will emerge to ensure that the benefits of AI are shared with the people whose work makes these models possible. Without deliberate action, the “death knell” may become a permanent silence for many communities that rely on local news for information, accountability, and identity.
Source: Windows Central News