Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteWindows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallThe New York Times’ lawsuit against OpenAI and Microsoft mattered because it put three questions in one case: whether copying journalism to train AI is fair use, whether a chatbot can infringe by reproducing protected passages, and whether AI answers can undercut the markets that fund reporting. Filed on December 27, 2023, the case was a defining copyright dispute in 2024—but it did not produce a final ruling that year. An April 4, 2025 decision dismissed some claims and allowed important copyright claims to continue.
What the Times alleged
The Times sued Microsoft and OpenAI in federal court in the Southern District of New York on December 27, 2023. Its complaint alleged that the companies copied millions of Times works in developing AI products, and that those products could reproduce or closely mimic the newspaper’s journalism. The Times asserted direct, contributory and vicarious copyright infringement, along with Digital Millennium Copyright Act, unfair-competition and trademark-dilution claims. These were allegations, not findings. Read the complaint.
| # | Preview | Product | Price | |
|---|---|---|---|---|
| 1 |
|
Copyright Law | $148.04 | Buy on Amazon |
| 2 |
|
Copyright Law: Cases and Materials (v8.0) | $21.70 | Buy on Amazon |
| 3 |
|
Copyright Law of the United States: and Related Laws Contained in Title 17 of the United States Code | $10.32 | Buy on Amazon |
| 4 |
|
Copyright Law in a Nutshell | $65.00 | Buy on Amazon |
| 5 |
|
Copyright Handbook, The: What Every Writer Needs to Know | $37.99 | Buy on Amazon |
The complaint’s theory was not simply that a model had encountered news on the internet. The Times argued that its reporting was used in commercial products that could provide information in competition with the newspaper and weaken its subscription, advertising, licensing, affiliate and other business opportunities. The central dispute therefore concerned both the copying itself and what the resulting products did with the material.
Two copyright questions: training and outputs
Copies made to prepare or train models
Training raises the question of whether making and using copies of copyrighted articles to build a model is fair use. That analysis is distinct from whether the model later produces infringing text. Content may be acquired, stored, preprocessed, used in training or fine-tuning, and retained in ways that raise different factual and legal questions.
#1 Best Overall
Text delivered to users
The Times also alleged that particular prompts produced long passages resembling or reproducing its articles. If a system gives a user protected expression from a specific work, that output can raise a separate infringement question from the legality of training copies. A brief quotation, a factual summary and a substantial near-verbatim reproduction are not interchangeable examples.
The Times’ complaint offered examples intended to show more than a model’s general ability to discuss news. OpenAI responded that the Times used deliberately engineered prompts to elicit unusual verbatim outputs, which it said did not reflect ordinary use. That is OpenAI’s position, not a court finding. A prompt designed to extract memorized text may affect how an example is assessed, but it does not by itself settle whether the underlying training was lawful. OpenAI’s public response.
Why fair use made the case difficult
U.S. fair use is assessed through four statutory factors. No one factor automatically decides the issue, and the analysis depends on the particular use and evidence.
- Purpose and character. OpenAI’s defense emphasizes that a model transforms training material into a system that can generate responses. The Times can point to commercial products that may answer questions in a way that competes with the original reporting. Commerciality matters, but it does not alone resolve fair use.
- Nature of the work. News contains facts, which copyright does not protect as such. But a reporter’s wording, analysis, selection and arrangement, investigative work and explanatory writing can be protected expression.
- Amount and substantiality. The alleged scale of training copies and the amount of a particular article reproduced in an answer are separate issues. A claim about extensive copying for training does not make every output infringing; nor does the fact that an answer is short necessarily mean it takes an insubstantial part of a work.
- Effect on actual or potential markets. This was the commercial center of the fight. The Times could argue that answers substitute for reading or licensing its reporting and impair a market for authorized AI access. OpenAI could argue that model training is transformative and that publishers should not control every system that learns from publicly accessible material. Evidence about subscriptions, traffic, licensing and product use would matter to the competing claims.
That framework is why neither “AI training is fair use” nor “AI training is infringement” is an adequate general conclusion. The answer can depend on what was copied, how it was used, what the product outputs, the market evidence and the specific claims before a court.
Recommended Free Tools
Why the Times was a consequential plaintiff
The Times brought a large archive of professionally produced journalism, a subscription and paywall business, and substantial capacity to litigate against well-funded technology companies. Those features made the claimed economic stakes concrete: the newspaper could argue that the same reporting it finances might be delivered by an AI product without a reader visiting or subscribing to the source.
The case also had implications beyond one newspaper. A result could influence bargaining positions for other publishers and for authors, image owners, software developers and database operators. But a district-court ruling would be tied to its facts and claims; it would not automatically settle every dispute over every kind of AI training.
Rank #3
Why Microsoft was more than a bystander
The complaint connected Microsoft to investment, infrastructure, integration and the commercial distribution of products using OpenAI technology, including GPT-based services. It alleged direct, contributory and vicarious infringement theories against the defendants. Those theories require proof of their respective elements; investment in OpenAI alone would not establish Microsoft’s liability.
The distinction matters because OpenAI’s role in developing and operating models is not identical to Microsoft’s role in infrastructure and product distribution. The Times had to connect each defendant to the alleged conduct under the claims it brought.
Do these 3 things before closing this tab:
1Scan for outdated or missing drivers - takes under a minute2Repair Windows errors before they cause bigger problems3Fix the driver behind crashes, sound loss and screen glitchesThe business stakes for journalism
Publishers earn revenue through more than subscriptions. Advertising, syndication, archive and data licensing, search referrals, affiliate commerce and authorized AI partnerships may all depend on control over reporting and access to readers. The Times alleged that AI-generated answers could use its work while sending fewer users to its site, weakening both audience relationships and its leverage in future licensing negotiations. That was a market-harm theory, not an established finding of damages.
Rank #4
The counterargument is that AI products might also help readers discover sources or create new distribution channels. Whether the net effect is substitution or expansion is an empirical question: it would require evidence about traffic, subscriptions, licensing and actual product use, not assumptions based only on the existence of a chatbot.
Why licensing was central—and not simple
OpenAI said it had discussed a partnership with the Times involving real-time display, attribution and access to reporting before the lawsuit. The breakdown of those discussions made licensing a visible alternative to litigation, but a license does not resolve every practical question.
- What is being licensed? Training, retrieval, real-time display and quoting may need different terms.
- How is payment measured? Possible bases include corpus size, tokens, users, revenue or outputs; the appropriate measure is not self-evident.
- How can use be audited? Publishers may seek ways to verify what material a system uses and how it is presented.
- What happens after a license ends? The parties would need to address retained models, retraining and ongoing access.
- Who can negotiate? Large publishers may have leverage that smaller outlets and independent creators lack.
Those questions explain why the likely policy landscape was not simply “court ruling or deal.” A hybrid could permit some training while requiring permission for high-value retrieval, verbatim output or real-time display. Whether copyright law supports those lines is a legal question; whether the industry can negotiate workable terms is a business one.
Best Value
What evidence could change the analysis
The dispute’s outcome would depend in part on evidence that is difficult to see from a public chatbot demonstration. Relevant questions include what materials were acquired and copied, how training and retrieval occurred, whether specific outputs reproduce protected expression, and whether products substitute for or direct audiences toward the publisher.
- Facts and ideas are not the same as copied article expression.
- Broad imitation of a writing style is different from reproducing protected wording, though the issues can overlap.
- A retrieval system that displays source text may present a different case from a model generating a general answer from learned patterns.
- Public availability online does not itself mean content is free of copyright restrictions. Technical controls and paywalls may be relevant context, but neither automatically answers the fair-use question.
- Output safeguards could reduce verbatim reproduction claims without resolving whether the original training copies were lawful.
What happened in 2024—and what the later ruling changed
The complaint was filed at the end of 2023, and 2024 brought pleadings, motions, discovery disputes and related publisher litigation rather than a final merits decision. OpenAI moved to dismiss in February 2024, arguing, among other things, that some claims were time-barred and that the unfair-competition theory was defective or preempted. OpenAI’s motion to dismiss.
A November 2024 discovery dispute illustrates how technical evidence became part of the fight. In a November 22 filing, OpenAI said a machine configuration change during inspection removed folder structure and file names from a temporary cache drive, while disputing that evidence had been destroyed. That was OpenAI’s account in a litigation filing, not a finding that either side acted improperly. The filing.
On April 4, 2025, the district court dismissed the Times’ common-law unfair-competition-by-misappropriation claim and certain DMCA claims, rejected some limitations-period arguments, and allowed important direct and contributory copyright claims to continue. The ruling narrowed the case; it did not clear OpenAI or decide the ultimate copyright merits. Read the opinion.
Free tools Windows power users keep installed
One-click scans. No signup required.
What the case could—and could not—decide
The case was worth watching because it put the economics of reporting and the legal treatment of commercial AI in the same courtroom. Its most consequential questions were whether training copies qualify as fair use, when a generated answer crosses into copying protected expression, and what evidence can establish harm to existing or emerging markets.
It could not, by itself, set a universal rule for every model, dataset or publisher. Training, retrieval and output may be treated differently; factual summaries differ from substantial reproduction; and any court ruling would apply through the claims and record before it. The lawsuit’s broader influence could also come from discovery demands, output controls, settlement terms and licensing practices adopted before a final judgment.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
