The New York Times Sues OpenAI and Microsoft for $20 Million! How Will News Copyright and Journalists’ Futures Evolve in the AI Era?

As artificial intelligence (AI) rapidly permeates the news industry, The New York Times’ copyright lawsuit against OpenAI and Microsoft (MSFT-US) has evolved from a single legal case into an industry-wide war over how tech companies obtain news content and how media organizations sustain their business models.

As of the end of July 2026, The New York Times has spent over $20 million on this lawsuit, and if the litigation continues, total costs could rise to $40 million.

For most news organizations, this amount could sustain an editorial department for several years, but The New York Times has no intention of ending the fight.

A.G. Sulzberger, publisher of The New York Times, believes this lawsuit is not merely about whether the newspaper can receive compensation, but whether technology companies can first take content that news organizations have spent years investing massive human and financial resources to produce, and then decide whether and how much to pay under the guise of 'technological innovation.'

He emphasizes that The New York Times is not opposed to AI, but to a business model that is emerging: tech companies are willing to pay high costs for talent, chips, data centers, and electricity, yet treat news, books, music, and films as freely available 'data.'

From Failed Negotiations to Courtroom Battle

This legal conflict did not originally begin in court.

In April 2023, The New York Times began negotiating with OpenAI and Microsoft to reach an agreement on the use of news content. At the time, ChatGPT had been launched for less than six months, and the publishing industry was gradually realizing that generative AI could not only change content creation but also become an alternative gateway to news websites.

However, the two sides ultimately failed to reach an agreement.

On December 27, 2023, The New York Times formally filed a lawsuit against OpenAI and Microsoft in the U.S. District Court for the Southern District of New York, accusing the two companies of unauthorized reproduction of millions of New York Times articles to train the GPT series of large language models. This marked the first time a major news organization had sued both OpenAI and Microsoft simultaneously.

The New York Times’ allegations focus on three main areas.

First, the model training phase. The newspaper argues that OpenAI copied its articles during the collection, organization, and input of training data, and that some of this content was subsequently 'remembered' by the model, potentially constituting unauthorized reproduction.

Second, the model output phase. The New York Times cited reports such as its Pulitzer Prize-winning investigation into the New York taxi license industry and 'Snow Fall: The Avalanche at Tunnel Creek' in its complaint, asserting that ChatGPT can generate content highly similar to, or even nearly verbatim reproductions of, the original articles.

Microsoft’s Browse with Bing service was also accused. It was alleged that the service directly compiled recommendations from Wirecutter, The New York Times’ consumer product review site, retaining the recommendations but removing original links and shopping links that generate affiliate revenue for Wirecutter.

Third, the substitution effect.

Previously, readers typically found news through search engines and clicked through to The New York Times’ website. Now, ChatGPT and Microsoft Copilot can directly provide answers, potentially reducing visits to the original articles and putting media at risk of losing traffic, subscriptions, advertising, and content licensing revenue.

The New York Times also believes that AI-generated misinformation could be mistaken as originating from the newspaper, thereby damaging the brand and credibility built over decades.

As for Microsoft, The New York Times did not name it as a defendant solely because of its investment in OpenAI. Microsoft provides funding and computing resources to OpenAI and integrates the related models into products such as Bing search, Copilot, and Azure. The New York Times therefore argues that Microsoft, having participated in building the technological system and using the models for commercial services, should bear corresponding responsibility.

The New York Times has not specified a particular damages amount but claims that OpenAI and Microsoft could face billions of dollars in statutory damages and actual losses. It also demands the destruction of training data and related models containing its copyrighted works.

If the court ultimately supports this demand, the impact will extend beyond compensation and could reshape how the entire AI industry acquires training data.

The New York Times Is Not Anti-AI—It Demands Prior Authorization for Content Use

In response to the lawsuit, OpenAI described The New York Times’ actions as hindering the development of new technology.

In a filing submitted in February 2024, OpenAI accused The New York Times of conducting tens of thousands of attempts on ChatGPT to obtain the model’s reproduction results cited in the complaint, using article fragments and special prompt words to induce model outputs. OpenAI argued that this was not how typical users normally use the product.

Microsoft, meanwhile, characterized The New York Times’ portrayal of AI’s impact on journalism as a 'doomsday scenario,' comparing it to the film industry’s past opposition to video recorders. It argued that while new technologies may change content consumption, they should not automatically be deemed infringing.

The core of both companies’ defense is 'fair use.'

OpenAI and Microsoft argue that large language models are learning language patterns, not creating a database for users to search original texts. Occasional reproduction of articles is a technical issue to be resolved, not evidence that the entire training process is illegal.

However, The New York Times has not chosen to reject AI entirely. Sulzberger stated that news organizations do not need to rush to be the first adopters of new technology. A more prudent approach is to observe, learn, and experiment while ensuring that journalists, editors, and news organizations remain responsible for the final content.

Internally, The New York Times is currently using AI more for assisting reporting rather than directly replacing journalists in writing, editing, and publishing.

For example, in an investigation into craters in the Gaza Strip, AI helped reporters filter thousands of potentially valuable images from millions of photos, but the final identification, verification, and judgment were still made by journalists. Reporters involved in the story described AI as a 'filter' that narrows the search scope but cannot replace human accountability in journalism.

This is Sulzberger’s core stance on AI: regardless of the tools used, the final responsibility for content published by The New York Times must lie with humans. AI can help journalists find clues, but it should not decide what is true or what content should be published.

Litigation and Licensing in Parallel—The New York Times Sets Red Lines for Cooperation

The New York Times is not refusing to cooperate with tech companies.

In May 2025, the newspaper reached its first generative AI content licensing agreement with Amazon (AMZN-US). Amazon can display summaries of content from The New York Times, NYT Cooking, and The Athletic in products like Alexa and can use the content to train its own foundational models.

The two parties did not officially disclose the transaction amount, but The Wall Street Journal later reported that Amazon pays approximately $20–25 million annually in licensing fees.

This creates a notable contrast: The New York Times has already spent over $20 million litigating against OpenAI and Microsoft, while simultaneously earning a roughly equivalent annual income from Amazon through licensing agreements.

Analysts point out that the real difference between the two is not whether AI is used, but whether authorization is obtained beforehand and whether news content is recognized as having value.

In December 2025, The New York Times sued AI search company Perplexity, accusing it of unauthorized reproduction and display of content, including paid articles.

These actions show that The New York Times is attempting to draw a line in its relationship with Silicon Valley: companies willing to pay and accept licensing terms can become partners; those that use content first and negotiate later may become defendants.

The Lawsuit Remains Unresolved—Both Sides Battle Over Evidence

On April 4, 2025, U.S. Federal District Judge Sidney H. Stein, presiding over the case, issued the first major ruling. The court allowed The New York Times’ primary copyright claims to proceed, rejecting OpenAI’s argument that some early training activities were beyond the statute of limitations and not dismissing the main allegations related to model output and user copying.

However, this does not mean The New York Times has won. The court also dismissed claims based on common law unfair competition and certain claims under the Digital Millennium Copyright Act (DMCA). More importantly, the judge did not rule on whether OpenAI’s model training constitutes fair use, only determining that The New York Times’ core factual allegations are sufficient to proceed to the discovery phase.

Since then, both sides have engaged in a battle over evidence. In May 2025, the judge ordered OpenAI to preserve ChatGPT output records that might otherwise be deleted, including conversations deleted by users. OpenAI argued this could violate its privacy commitments to users.