Background
The New York Times alleges that Microsoft and OpenAI copied and used millions of Times works to train GPT models and related products, including ChatGPT, Bing Chat and Browse with Bing, and that those systems compete with the Times by reproducing, paraphrasing, and substituting for its journalism and Wirecutter content without permission or payment [ECF No. 1; ECF No. 170; ECF No. 585]. The pleading evolved from the original complaint to a first amended complaint and then to a second amended complaint, while the case was later folded into the SDNY OpenAI copyright MDL, where the court set consolidated case-management and leadership structures for the broader litigation [ECF No. 156; ECF No. 574]. At the level reviewed here, the court has not decided liability or fair use, but it has already ruled on the main motions to dismiss: in March 2025 it allowed the core copyright case to continue, including the contributory copyright theories and the challenged older direct-infringement allegations, while dismissing the common-law misappropriation claim and narrowing the DMCA claims [ECF No. 485]. The matter is now beyond pleadings and into consolidated discovery, expert scheduling, and MDL case management [ECF No. 689].
AI interaction
The AI-specific core of the case is that the Times says the defendants used Times journalism as high-value training material for large language models, encoded Times content into model parameters, and then generated outputs that could reproduce or closely paraphrase Times reporting, including through synthetic search responses, Wirecutter-style recommendations, and false attributions or “hallucinations” under the Times name [ECF No. 585]. On the defence side, OpenAI and Microsoft moved to dismiss selected claims and framed model training as lawful and technologically transformative [ECF Nos. 52, 65]. At this stage, however, the court did not adopt a conclusive legal view that AI training is lawful or unlawful as such. Instead, its ruling shows a more limited but important position: the court treated the main AI-copyright dispute as sufficiently pleaded and too fact-dependent to be resolved away on a motion to dismiss, while cutting back only certain non-core theories such as misappropriation and parts of the DMCA claims [ECF No. 485]. So the present AI significance of the case is procedural but substantial: the court has allowed the central copyright questions around training, model outputs, and AI-generated substitutes for news content to move forward into full factual development rather than disposing of them at the threshold stage.
Notes:
- The case remains active in the MDL.
- In March 2026, the court substituted OpenAI Group PBC in place of OpenAI Holdings, LLC [ECF No. 1342], and an amended Rule 7.1 disclosure stated that OpenAI, Inc. had changed its name to OpenAI Foundation in October 2025 and identified Microsoft Corporation and SoftBank Group Corporation as publicly held corporations with 10% or greater financial interests in OpenAI Group PBC on an as-converted basis [ECF No. 1353].
- A further March 2026 scheduling order also set another confidential discovery settlement conference and required proposals for further briefing and status-conference dates, confirming that the case is still in a live discovery-management phase rather than near final adjudication [ECF No. 1347].