Altorise is a marketing accelerator and growth engine. We work with ambitious founders and investors to grow breakthrough ideas. Using a combination of marketing, design, and technology skills we engineer scalable growth. Reach us at hi@altorise.com | +91 93268 44428 | @altorise

Altorise

The New York Times Company v. Microsoft Corporation Court Filing December 27, 2023 is part of . You can jump to any part in this filing . This is part 10 of 27. HackerNoon’s Legal PDF Series here IV. FACTUAL ALLEGATIONS C. Defendants’ Unauthorized Use and Copying of Times Content 82. Microsoft and OpenAI created and distributed reproductions of The Times’s content in several, independent ways in the course of training their LLMs and operating the products that incorporate them. 1. Unauthorized Reproduction of Times Works During GPT Model Training 83. Defendants’ GPT models are a family of LLMs, the first of which was introduced in 2018, followed by GPT-2 in 2019, GPT-3 in 2020, GPT-3.5 in 2022, and GPT-4 in 2023. The “chat” style LLMs, GPT-3.5 and GPT-4, were developed in two stages. First, a transformer model was pre-trained on a very large amount of data. Second, the model was “fine-tuned” on a much smaller supervised dataset in order to help the model solve specific tasks. 84. The pre-training step involved collecting and storing text content to create training datasets and processing that content through the GPT models. While OpenAI did not release the trained versions of GPT-2 onward, “[d]ue to [OpenAI’s] concerns about malicious applications of the technology,” OpenAI has published general information about its pre-training process for the GPT models.[12] 85. GPT-2 includes 1.5 billion parameters, which was a 10X scale up of GPT.[13] The training dataset for GPT-2 includes an internal corpus OpenAI built called “WebText,” which includes “the text contents of 45 million links posted by users of the ‘Reddit’ social network.”[14] The contents of the WebText dataset were created as a “new web scrape which emphasizes document quality.”[15] The WebText dataset contains a staggering amount of scraped content from The Times. For example, the NYTimes.com domain is one of the “top 15 domains by volume” in the WebText dataset,[16] and is listed as the 5th “top domain” in the WebText dataset with 333,160 entries.[17] 86. GPT-3 includes 175 billion parameters and was trained on the datasets listed in the table below.[18] 87. One of these datasets, WebText2, was created to prioritize high value content. Like the original WebText, it is composed of popular outbound links from Reddit. As shown in the table above, the WebText2 corpus was weighted 22% in the training mix for GPT-3 despite constituting less than 4% of the total tokens in the training mix. Times content—a total of 209,707 unique URLs—accounts for 1.23% of all sources listed in OpenWebText2, an open-source re-creation of the WebText2 dataset used in training GPT-3. Like the original WebText, OpenAI describes WebText2 as a “high-quality” dataset that is “an expanded version of the WebText dataset … collected by scraping links over a longer period of time.”[19] 88. The most highly weighted dataset in GPT-3, Common Crawl, is a “copy of the Internet” made available by an eponymous 501(c)(3) organization run by wealthy venture capital investors.[20] The domain www.nytimes.com is the most highly represented proprietary source (and the third overall behind only Wikipedia and a database of U.S. patent documents) represented in a filtered English-language subset of a 2019 snapshot of Common Crawl, accounting for 100 million tokens (basic units of text): [21] 89. The Common Crawl dataset includes at least 16 million unique records of content from The Times across News, Cooking, Wirecutter, and The Athletic, and more than 66 million total records of content from the Times. 90. Critically, OpenAI admits that “datasets we view as higher-quality are sampled more frequently” during training.[22] Accordingly, by OpenAI’s own admission, high-quality content, including content from The The Times, was more important and valuable for training the GPT models as compared to content taken from other, lower-quality sources. 91. While OpenAI has not released much information about GPT-4, experts suspect that GPT-4 includes 1.8 trillion parameters, which is over 10X larger than GPT-3, and was trained on approximately 13 trillion tokens.[23] The training set for GPT-3, GPT-3.5, and GPT-4 was comprised of 45 terabytes of data—the equivalent of a Microsoft Word document that is over 3.7 billion pages long. [24] Between the Common Crawl, WebText, and WebText2 datasets, the Defendants likely used millions of Times-owned works in full in order to train the GPT models. 92. Defendants repeatedly copied this mass of Times copyrighted content, without any license or other compensation to The Times. As part of training the GPT models, Microsoft and OpenAI collaborated to develop a complex, bespoke supercomputing system to house and reproduce copies of the training dataset, including copies of The Times-owned content. Millions of Times Works were copied and ingested—multiple times—for the purpose of “training” Defendants’ GPT models. 93. Upon information and belief, Microsoft and OpenAI acted jointly in the large-scale copying of The Times’s material involved in generating the GPT models programmed to accurately mimic The Times’s content and writers. Microsoft and OpenAI collaborated in designing the GPT models, selecting the training datasets, and supervising the training process. As Mr. Nadella stated: So, there are a lot of, I call it, product design choices one gets tomake when you think about AI and AI safety. Then, let’s come at itthe other way. You have to take real care of the pretrained data because models are trained on pretrained data. What’s the quality,the provenance of that pretrained data? That’s a place where we’vedone a lot of work.[25] 94. To the extent that Microsoft did not select the works used to train the GPT models, it acted in self-described “partnership” with OpenAI respecting that selection, knew or was willfully blind to the identity of the selected works by virtue of its knowledge of the nature and identity of the training corpuses and selection criteria employed by OpenAI, and/or had the right and ability to prevent OpenAI from using any particular work for training by virtue of its physical control of the supercomputer it developed for that purpose and its legal and financial influence over the OpenAI Defendants. 95. Upon information and belief, Microsoft and OpenAI continue to create unauthorized copies of Times Works in the form of synthetic search results returned by their Bing Chat and Browse with Bing products. Microsoft actively gathers copies of the Times Works used to generate such results in the process of crawling the web to create the index for its Bing search engine. 96. On information and belief, Microsoft and OpenAI are currently or will imminently commence making additional copies of Times Works to train and/or fine-tune the next-generation GPT-5 LLM. 97. Defendants’ large-scale commercial exploitation of Times content is not licensed, nor have Defendants received permission from The Times to copy and use its works to build their GenAI tools. Continue Reading . Here [12] OpenAI, Better Language Models and Their Implications, OPENAI (Feb. 14, 2019), https://openai.com/research/better-language-models. [13] Id. [14] GPT-2 Model Card, GITHUB (Nov. 2019), https://github.com/openai/gpt-2/blob/master/model_card.md. [15] RADFORD ET AL., LANGUAGE MODELS ARE UNSUPERVISED MULTITASK LEARNERS 3 (2018), https://d4mucfpksywv.cloudfront.net/better-language-models/language-models.pdf. [16] GPT-2 Model Card, supra note 14. [17] GPT-2 / domains.txt, GITHUB, https://github.com/openai/gpt-2/blob/master/domains.txt (last visited Dec. 21, 2023). [18] BROWN ET AL., LANGUAGE MODELS ARE FEW-SHOT LEARNERS 9 (2020), https://arxiv.org/pdf/2005.14165.pdf. [19] Id. at 8. [20] COMMON CRAWL, https://commoncrawl.org/ (last visited Dec. 21, 2023). [21] DODGE ET AL., DOCUMENTING LARGE WEBTEXT CORPORA: A CASE STUDY ON THE COLOSSAL CLEAN CRAWLED CORPUS (2021), https://arxiv.org/abs/2104.08758. [22] BROWN ET AL., supra note 18. [23] Maximilian Schreiner, GPT-4 Architecture, Datasets, Costs and More Leaked, THE DECODER (July 11, 2023), https://the-decoder.com/gpt-4-architecture-datasets-costs-and-more-leaked/. [24] Kindra Cooper, OpenAI GPT-3: Everything You Need to Know [Updated], SPRINGBOARD (Sept. 27, 2023), https://www.springboard.com/blog/data-science/machine-learning-gpt-3-open-ai/. [25] Nilay Patel, Microsoft Thinks AI Can Beat Google at Search — CEO Satya Nadella Explains Why, THE VERGE (Feb. 7, 2023), https://www.theverge.com/23589994/microsoft-ceo-satya-nadella-bing-chatgpt-googlesearch-ai. About HackerNoon Legal PDF Series: We bring you the most important technical and insightful public domain court case filings. This court case 1:23-cv-11195 retrieved on December 29, 2023, from is part of the public domain. The court-created documents are works of the federal government, and under copyright law, are automatically placed in the public domain and may be shared without legal restriction. nycto-assets.nytimes.com

The Times v. Microsoft/OpenAI: Unauthorized Reproduction of Times Works In GPT Model Training (10)

টাইমস বনাম মাইক্রোসফ্ট/ওপেনএআই: জিপিটি মডেল প্রশিক্ষণে টাইমসের অননুমোদিত পুনরুত্পাদন (10)

The Times vs. Microsoft/OpenAI: Unbefugte Reproduktion von Times Works im GPT-Modelltraining (10)

The Times contra Microsoft/OpenAI: la reproducción no autorizada de Times funciona en el entrenamiento del modelo GPT (10)

The Times contre Microsoft/OpenAI : reproduction non autorisée de Times Works dans la formation sur le modèle GPT (10)

द टाइम्स बनाम माइक्रोसॉफ्ट/ओपनएआई: जीपीटी मॉडल प्रशिक्षण में टाइम्स वर्क्स का अनधिकृत पुनरुत्पादन (10)

Times 対 Microsoft/OpenAI: GPT モデル トレーニングにおける Times Works の無断転載 (10)

The Times 대 Microsoft/OpenAI: GPT 모델 교육에서 시간의 무단 복제가 작동합니다(10)

The Times x Microsoft/OpenAI: a reprodução não autorizada do Times funciona no treinamento do modelo GPT (10)

The Times против Microsoft/OpenAI: несанкционированное воспроизведение времен работает при обучении модели GPT (10)

The Times - Microsoft/OpenAI: GPT Model Eğitiminde Times Çalışmalarının İzinsiz Çoğaltılması (10)

The Times v. Microsoft/OpenAI: Sao chép trái phép Times Works trong đào tạo mô hình GPT (10)

《泰晤士报》诉微软/OpenAI：未经授权复制《泰晤士报》在 GPT 模型训练中有效 (10)

The New York Times Company v. Microsoft Corporation Court Filing December 27, 2023 is part of . You can jump to any part in this filing . This is part 11 of 27. HackerNoon’s Legal PDF Series here IV. FACTUAL ALLEGATIONS C. Defendants’ Unauthorized Use and Copying of Times Content 2. Embodiment of Unauthorized Reproductions and Derivatives of Times Works in GPT Models 98. As further evidence of being trained using unauthorized copies of Times Works, the GPT LLMs themselves have “memorized” copies of many of those same works encoded into their parameters. As shown below and in Exhibit J, the current GPT-4 LLM will output near-verbatim copies of significant portions of Times Works when prompted to do so. Such memorized examples constitute unauthorized copies or derivative works of the Times Works used to train the model. 99. For example, in 2019, The Times published a Pulitzer-prize winning, five-part series on predatory lending in New York City’s taxi industry. The 18-month investigation included 600 interviews, more than 100 records requests, large-scale data analysis, and the review of thousands of pages of internal bank records and other documents, and ultimately led to criminal probes and the enactment of new laws to prevent future abuse. OpenAI had no role in the creation of this content, yet with minimal prompting, will recite large portions of it verbatim:[26] 100. Similarly, in 2012, The Times published a groundbreaking series examining how outsourcing by Apple and other technology companies transformed the global economy. The series was the product of an enormous effort across three continents. Reporting this story was especially challenging because The Times was repeatedly denied both interviews and access. The Times contacted hundreds of current and former Apple executives, and ultimately secured information from more than six dozen Apple insiders. Again, GPT-4 copied this content and can recite large portions of it verbatim:[27] 101. Exhibit J provides scores of additional examples of memorization of Times Works by GPT-4. Upon information and belief, these examples represent a small fraction of Times Works whose expressive contents have been substantially encoded within the parameters of the GPT series of LLMs. Each of those LLMs thus embodies many unauthorized copies or derivatives of Times Works. Continue Reading . Here [26] For original article, see Brian M. Rosenthal, As Thousands of Taxi Drivers Were Trapped in Loans, Top Officials Counted the Money, N.Y. TIMES (May 19, 2019), https://www.nytimes.com/2019/05/19/nyregion/taximedallions.html. [27] For original article, see Charles Duhigg & Keith Bradsher, How the U.S. Lost Out on iPhone Work, N.Y. TIMES (Jan. 21, 2012), https://www.nytimes.com/2012/01/22/business/apple-america-and-a-squeezed-middleclass.html. About HackerNoon Legal PDF Series: We bring you the most important technical and insightful public domain court case filings. This court case 1:23-cv-11195 retrieved on December 29, 2023, from is part of the public domain. The court-created documents are works of the federal government, and under copyright law, are automatically placed in the public domain and may be shared without legal restriction. nycto-assets.nytimes.com

The Times v. Microsoft/OpenAI: Unauthorized Reproductions of Times Works in GPT Models (11)

দ্য টাইমস বনাম মাইক্রোসফট/ওপেনএআই: জিপিটি মডেলে টাইমস ওয়ার্কসের অননুমোদিত পুনরুৎপাদন (11)

The Times vs. Microsoft/OpenAI: Unautorisierte Reproduktionen von Times-Werken in GPT-Modellen (11)

The Times contra Microsoft/OpenAI: Reproducciones no autorizadas de obras de Times en modelos GPT (11)

The Times contre Microsoft/OpenAI : reproductions non autorisées des œuvres du Times dans des modèles GPT (11)

द टाइम्स बनाम माइक्रोसॉफ्ट/ओपनएआई: जीपीटी मॉडल में टाइम्स वर्क्स का अनधिकृत पुनरुत्पादन (11)

タイムズ対マイクロソフト/OpenAI: GPT モデルでの Times Works の不正複製 (11)

The Times 대 Microsoft/OpenAI: GPT 모델에서 작동하는 Times의 무단 복제(11)

The Times v. Microsoft/OpenAI: Reproduções não autorizadas de Times Works em modelos GPT (11)

The Times против Microsoft/OpenAI: несанкционированное воспроизведение Times работает в моделях GPT (11)

The Times - Microsoft/OpenAI: Times Works'ün GPT Modellerinde İzinsiz Çoğaltılması (11)

The Times kiện Microsoft/OpenAI: Sao chép trái phép Times Works trong Mô hình GPT (11)

《泰晤士报》诉 Microsoft/OpenAI：未经授权复制《泰晤士报》在 GPT 模型中的工作 (11)

The New York Times Company v. Microsoft Corporation Court Filing December 27, 2023 is part of . You can jump to any part in this filing . This is part 12 of 27. HackerNoon’s Legal PDF Series here IV. FACTUAL ALLEGATIONS C. Defendants’ Unauthorized Use and Copying of Times Content 3. Unauthorized Public Display of Times Works in GPT Product Outputs 102. Defendants directly engaged in the unauthorized public display of Times Works as part of generative output provided by their products built on the GPT models. Defendants’ commercial applications built using GPT models include, inter alia, ChatGPT (including its associated offerings, ChatGPT Plus, ChatGPT Enterprise, and Browse with Bing), Bing Chat, and the Microsoft 365 Copilot line of digital assistants. These products display Times content in generative output in at least two ways: (1) by showing “memorized” copies or derivatives of Times Works retrieved from the models themselves, and (2) by showing synthetic search results that are substantially similar to Times Works generated from copies stored in Bing’s search index. 103. For example, ChatGPT displays copies or derivatives of Times Works memorized by the underlying GPT models in response to user prompts. Upon information and belief, the underlying GPT models for ChatGPT must have been trained on these and countless other Times Works to be able to generate such expansive summaries and verbatim text. 104. Below, ChatGPT quotes part of the 2012 Pulitzer Prize-winning New York Times article “Snow Fall: The Avalanche at Tunnel Creek,” which was generated in response to a prompt complaining about being “paywalled out” of the article:[28] 105. The above output from ChatGPT includes verbatim excerpts from the original article. The copied article text is highlighted in red below: 106. Below, ChatGPT purports to quote former Times restaurant critic Pete Wells’s 2012 review of Guy Fieri’s American Kitchen & Bar, an article that has been described as a viral sensation: [29] 107. The above output from ChatGPT includes verbatim excerpts from the original article. The copied article text is highlighted in red below: Continue Reading . Here [28] For original article, see John Branch, Snow Fall: The Avalanche at Tunnel Creek, N.Y. TIMES (Dec. 13, 2012), https://www.nytimes.com/projects/2012/snow-fall/index.html#/?part=tunnel-creek. [29] For original article, see Pete Wells, As Not Seen on TV, N.Y. TIMES (Nov. 13, 2012), https://www.nytimes.com/2012/11/14/dining/reviews/restaurant-review-guys-american-kitchen-bar-in-timessquare.html. About HackerNoon Legal PDF Series: We bring you the most important technical and insightful public domain court case filings. This court case 1:23-cv-11195 retrieved on December 29, 2023, from is part of the public domain. The court-created documents are works of the federal government, and under copyright law, are automatically placed in the public domain and may be shared without legal restriction. nycto-assets.nytimes.com

The Times v Microsoft/OpenAI: Unauthorized Public Display of Times Works in GPT Product Outputs (12)

টাইমস বনাম মাইক্রোসফ্ট/ওপেনএআই: জিপিটি প্রোডাক্ট আউটপুটে টাইমস ওয়ার্কসের অননুমোদিত পাবলিক ডিসপ্লে (12)

The Times vs. Microsoft/OpenAI: Unbefugte öffentliche Anzeige von Times-Werken in GPT-Produktausgaben (12)

The Times v Microsoft/OpenAI: La visualización pública no autorizada de Times funciona en resultados de productos GPT (12)

The Times contre Microsoft/OpenAI : affichage public non autorisé des horaires dans les résultats des produits GPT (12)

टाइम्स बनाम माइक्रोसॉफ्ट/ओपनएआई: जीपीटी उत्पाद आउटपुट में टाइम्स वर्क्स का अनधिकृत सार्वजनिक प्रदर्शन (12)

The Times v Microsoft/OpenAI: GPT 製品出力における Time Works の無許可の公開表示 (12)

The Times v Microsoft/OpenAI: GPT 제품 출력에서 시간의 무단 공개 표시가 작동합니다. (12)

The Times v Microsoft/OpenAI: exibição pública não autorizada de Times funciona em resultados de produtos GPT (12)

The Times против Microsoft/OpenAI: несанкционированное публичное отображение времени работает в выходных данных продукта GPT (12)

The Times - Microsoft/OpenAI: Times Works'ün GPT Ürün Çıktılarında İzinsiz Herkese Açık Gösterimi (12)

The Times v Microsoft/OpenAI: Hiển thị công khai trái phép các tác phẩm của Times trong đầu ra sản phẩm GPT (12)

《泰晤士报》诉 Microsoft/OpenAI：未经授权公开显示《泰晤士报》在 GPT 产品输出中有效 (12)

The New York Times Company v. Microsoft Corporation Court Filing December 27, 2023 is part of . You can jump to any part in this filing . This is part 13 of 27. HackerNoon’s Legal PDF Series here IV. FACTUAL ALLEGATIONS C. Defendants’ Unauthorized Use and Copying of Times Content 4. Unauthorized Retrieval and Dissemination of Current News 108. Synthetic search applications built on the GPT LLMs, including Bing Chat and Browse with Bing for ChatGPT, display extensive excerpts or paraphrases of the contents of search results, including Times content, that may not have been included in the model’s training set. The “grounding” technique employed by these products includes receiving a prompt from a user, copying Times content relating to the prompt from the internet, providing the prompt together with the copied Times content as additional context for the LLM, and having the LLM stitch together paraphrases or quotes from the copied Times content to create natural-language substitutes that serve the same informative purpose as the original. In some cases, Defendants’ models simply spit out several paragraphs of The Times’s articles. 109. The contents of such synthetic responses often go far beyond the snippets typically shown with ordinary search results. Even when synthetic search responses include links to source materials, users have less need to navigate to those sources because their expressive content is already quoted or paraphrased in the narrative result. Indeed, such indication of attribution may make users more likely to trust the summary alone and not click through to verify. 110. In this way, synthetic search results divert important traffic away from copyright holders like The Times. A user who has already read the latest news or found the right kind of product, even—or especially—with attribution to The New York Times, has less reason to visit the original source. 111. Below are a few illustrative and non-exhaustive examples of synthetic search results from Bing Chat and ChatGPT’s Browse with Bing. a) Examples of Synthetic Search Results from Bing Chat 112. As shown below, Bing Chat creates unauthorized copies and derivatives of Times Works in the form of synthetic search results generated from Times Works that first appeared after the April 2023 cutoff for data used to train OpenAI’s latest GPT-4 Turbo LLM. 30 The first includes a long quote from the October 2023 New York Times article “The Secrets Hamas knew about Israel’s Military”:[31] 113. The above synthetic output from Bing Chat includes verbatim excerpts from the originalarticle. The copied article text is highlighted in red below. 114. The synthetic output displays significantly more expressive content from the original article than what would traditionally be displayed in a Bing search result for the same article, as shown below. Unlike a traditional search result, the synthetic output also does not include a prominent hyperlink that sends users to The Times’s website. 115. A further example shows Bing Chat extensively reproducing text from the September 2023 New York Times article “To Experience Paris Up Close and Personal, Plunge Into a Public Pool”:[32] 116. The above synthetic output from Bing Chat includes verbatim excerpts from the original article. The copied article text is highlighted in red below. 117. The synthetic output displays significantly more expressive content from the original article than what would traditionally be displayed in a Bing search result for the same article, as shown below. Unlike a traditional search result, the synthetic output also does not include a prominent hyperlink that sends users to The Times’s website. b) Synthetic Search Results from ChatGPT Browse with Bing 118. The below examples show that ChatGPT’s Browse with Bing plug-in also outputs unauthorized copies and derivatives of copyrighted works from The Times in the form of synthetic search results generated from Times Works that first appeared after the April 2023 cutoff for data used to train OpenAI’s latest GPT-4 Turbo LLM. The first reproduces the first two paragraphs of the May 2023 New York Times article “The Precarious, Terrifying Hours After a Woman Was Shoved Into a Train”:[33] 119. The above synthetic output from ChatGPT with the Browse with Bing plugin includes verbatim excerpts from the original article. The copied article text is highlighted in red below 120. The synthetic output displays significantly more expressive content from the original article than what would traditionally be displayed in a Bing search result for the same article as shown below. Unlike a traditional search result, the synthetic output also does not include a prominent hyperlink that sends users to The Times’s website. 121. This example likewise shows Browse with Bing for ChatGPT reproducing the first two paragraphs of The New York Times article “Are the Hamptons Still Hip?” from May 2023.[34] 122. The above synthetic output from ChatGPT with the Browse with Bing plugin includes verbatim excerpts from the original article. The copied article text is highlighted in red below. 123. Again, the synthetic output displays significantly more expressive content from the original article than what would traditionally be displayed in a Bing search result for the same article, as shown below. Unlike a traditional search result, the synthetic output also does not include a prominent hyperlink that sends users to The Times’s website. Continue Reading . Here [31] For original article, see Patrick Kingsley & Ronen Bergman, The Secrets Hamas Knew About Israel’s Military, N.Y. TIMES (Oct. 13, 2023), https://www.nytimes.com/2023/10/13/world/middleeast/hamas-israel-attackgaza.html. [32] For original article, see Catherine Porter, To Experience Paris Up Close and Personal, Plunge Into a Public Pool, N.Y. TIMES (Sept. 3, 2023), https://www.nytimes.com/2023/09/03/world/europe/paris-franceswimming-pools.html. [33] For original content, see Hurubie Meko, The Precarious, Terrifying Hours After a Woman Was Shoved Into a Train, N.Y. TIMES (May 25, 2023), https://www.nytimes.com/2023/05/25/nyregion/subway-attack-womanshoved-manhattan.html. [34] For original article, see Anna Kodé, Are the Hamptons Still Hip?, N.Y. TIMES (May 26, 2023), https://www.nytimes.com/2023/05/26/realestate/hamptons-summer-housing-costs.html. About HackerNoon Legal PDF Series: We bring you the most important technical and insightful public domain court case filings. This court case 1:23-cv-11195 retrieved on December 29, 2023, from is part of the public domain. The court-created documents are works of the federal government, and under copyright law, are automatically placed in the public domain and may be shared without legal restriction. nycto-assets.nytimes.com

The Times v. Microsoft/OpenAI: Unauthorized Retrieval and Dissemination of Current News (13)

টাইমস বনাম মাইক্রোসফ্ট/ওপেনএআই: বর্তমান সংবাদের অননুমোদিত পুনরুদ্ধার এবং প্রচার (13)

The Times vs. Microsoft/OpenAI: Unbefugter Abruf und Verbreitung aktueller Nachrichten (13)

The Times contra Microsoft/OpenAI: recuperación y difusión no autorizada de noticias actuales (13)

The Times contre Microsoft/OpenAI : récupération et diffusion non autorisées de l'actualité (13)

द टाइम्स बनाम माइक्रोसॉफ्ट/ओपनएआई: वर्तमान समाचारों की अनधिकृत पुनर्प्राप्ति और प्रसार (13)

タイムズ対マイクロソフト/OpenAI: 最新ニュースの不正取得と配布 (13)

The Times 대 Microsoft/OpenAI: 최신 뉴스의 무단 검색 및 배포(13)

The Times v. Microsoft/OpenAI: recuperação e disseminação não autorizada de notícias atuais (13)

The Times против Microsoft/OpenAI: несанкционированное получение и распространение текущих новостей (13)

The Times - Microsoft/OpenAI: Güncel Haberlerin İzinsiz Alınması ve Yayılması (13)

The Times v. Microsoft/OpenAI: Truy xuất và phổ biến trái phép tin tức hiện tại (13)

《泰晤士报》诉 Microsoft/OpenAI：未经授权检索和传播时事新闻 (13)

The New York Times Company v. Microsoft Corporation Court Filing December 27, 2023 is part of . You can jump to any part in this filing . This is part 14 of 27. HackerNoon’s Legal PDF Series here IV. FACTUAL ALLEGATIONS C. Defendants’ Unauthorized Use and Copying of Times Content 5. Willful Infringement 124. Defendants’ unauthorized reproduction and display of Times Works is willful. Defendants were intimately involved in training, fine-tuning, and otherwise testing the GPT models. Defendants knew or should have known that these actions involved unauthorized copying of Times Works on a massive scale during training, resulted in the unauthorized encoding of huge numbers of such works in the models themselves, and would inevitably result in the unauthorized display of such works that the models had either memorized or would present to users in the form of synthetic search results. In fact, in late 2023 before his ouster and subsequent reinstatement as OpenAI’s CEO, Sam Altman reportedly clashed with OpenAI board member Helen Toner over a paper that Toner wrote criticizing the company over “safety and ethics issues related to the launches of ChatGPT and GPT-4, including regarding copyright issues.” 125. The Times specifically put Defendants on notice that these uses of Times Works were not authorized by placing copyright notices and linking to its terms of service (which contain, among other things, terms and conditions for the use of its works) on every page of its websites whose contents Defendants copied and displayed. Upon information and belief, Defendants intentionally removed such copyright management information (“CMI”) from Times Works in the process of preparing them to be used to train their models with the knowledge that such CMI would not be retained within the models or displayed when the models present unauthorized copies or derivatives of Times Works to users, and thereby facilitate or conceal their infringement. 126. Upon information and belief, Defendants were aware of many examples of copyright infringement after ChatGPT, Browse with Bing, and Bing Chat were released, some of which were widely publicized. In fact, after the release of ChatGPT and Bing Chat, The Times reached out to Defendants to inform them that their tools infringed its copyrighted works. Continue Reading . Here About HackerNoon Legal PDF Series: We bring you the most important technical and insightful public domain court case filings. This court case 1:23-cv-11195 retrieved on December 29, 2023, from is part of the public domain. The court-created documents are works of the federal government, and under copyright law, are automatically placed in the public domain and may be shared without legal restriction. nycto-assets.nytimes.com

The Times v. Microsoft/OpenAI: Defendants’ Unauthorized Use and Copying of Times Content (14)

টাইমস বনাম মাইক্রোসফ্ট/ওপেনএআই: বিবাদীদের অননুমোদিত ব্যবহার এবং টাইমস সামগ্রীর অনুলিপি (14)

The Times gegen Microsoft/OpenAI: Unbefugte Nutzung und Vervielfältigung von Times-Inhalten durch die Beklagten (14)

The Times contra Microsoft/OpenAI: Uso no autorizado y copia del contenido del Times por parte de los demandados (14)

The Times c. Microsoft/OpenAI : utilisation non autorisée et copie du contenu du Times par les défendeurs (14)

द टाइम्स बनाम माइक्रोसॉफ्ट/ओपनएआई: प्रतिवादियों द्वारा टाइम्स सामग्री का अनधिकृत उपयोग और नकल (14)

タイムズ対マイクロソフト/OpenAI: 被告によるタイムズコンテンツの不正使用とコピー (14)

The Times 대 Microsoft/OpenAI: 피고의 Times 콘텐츠 무단 사용 및 복사(14)

The Times v. Microsoft/OpenAI: Uso não autorizado e cópia do conteúdo do Times pelos réus (14)

The Times против Microsoft/OpenAI: Несанкционированное использование и копирование контента Times ответчиками (14)

The Times / Microsoft/OpenAI: Davalıların Times İçeriğinin İzinsiz Kullanımı ve Kopyalanması (14)

The Times v. Microsoft/OpenAI: Bị cáo sử dụng trái phép và sao chép nội dung của Times (14)

《泰晤士报》诉 Microsoft/OpenAI：被告未经授权使用和复制《泰晤士报》内容 (14)

We've looked at many different ways to create live streaming experiences with Amazon Interactive Video Service (Amazon IVS) here on this blog, but all of those examples so far have focused on public channels that can be viewed by anyone without any playback authorization to verify that the viewer is allowed to view the live stream. Many live streaming user generated content (UGC) platforms do not require any authorization to view a stream, but there are use cases where we want to limit our stream viewers to a subset of our audience. For example, maybe our content creators want to charge a subscription for the content that they produce. Or, perhaps there are licensing limitations to the stream content that prohibit distribution to certain demographics or geographic locations. To address use cases where the platform must enforce playback authorization, Amazon IVS supports private channels. These channels will restrict playback to viewers who have obtained a token that is signed with a private key and verified with a public key. In this post, we'll look at how to enable playback authorization, create a playback key pair, generate a token, and play a private stream. Creating a Private Channel The first step in this workflow is to create a private channel. Creating a private channel is very similar to creating a public channel - the only difference is a single setting when creating the channel to enable playback authorization. We can enable playback authorization via the , via the AWS CLI, or via any of the AWS SDKs. Amazon IVS Console Let's start by looking at how to create a private channel via the Amazon IVS Console. At the list of channels in the console, click . Create channel On the page, enter a channel name (#1), select (#2), and turn on (#3). Create channel Custom configuration Enable token-authorization requirement for video playback If you prefer to use the CLI to create a channel, you can add the option. --authorized $ aws ivs create-channel --name demo-private-channel --latency-mode LOW --type STANDARD --authorized In reality, you're probably going to use the AWS SDK to create your user's channels. In this case, you can refer to the AWS SDK documentation for your favorite language to determine the appropriate argument to set for playback. For example, here are the for for the AWS SDK for JavaScript v3 which state that we can set a boolean property when creating a channel. authorized docs CreateChannelCommandInput authorized Now that we have enabled playback authorization, if we were to broadcast to the channel and try to view the live playback without generating and passing a playback token, we will receive a playback error. Here's an example of playback using the Amazon IVS player SDK where we have added an listener to capture playback errors. ERROR const streamUrl = '[private channel Playback Url]';
const ivsPlayer = IVSPlayer.create();
ivsPlayer.attachHTMLVideoElement(document.getElementById('video-player'));
ivsPlayer.load(streamUrl);
ivsPlayer.play();

ivsPlayer.addEventListener(IVSPlayer.PlayerEventType.ERROR, (err) => {
  console.log(err);
}); Running this in a browser when pointed at a private channel will result in the following error logged to the console. {
  "type": "ErrorAuthorization",
  "code": 7,
  "source": "MasterPlaylist",
  "message": "invalid_playback_auth_token"
} This confirms that our channel is indeed private, and requires a to be generated and passed along with the to authorize playback. token Playback Url Creating a Playback Key Before we can generate a playback token, we will need to create a private/public key pair and upload it to the Amazon IVS console. This key pair will be used to generate a token that is signed with the private key, and verified with the public key, to authorize playback. JWT To manage Playback keys with the Amazon IVS Console, select in the left navigation sidebar. Playback keys We can either create a key with the console, or manually with OpenSSL offline on our local machine. To create a key pair directly in the console. For this, click on . Create playback key Give the key a name, and click . Create The public key of this pair is saved to your AWS account and will be used to verify any playback requests that contain a token signed with the private key. When using the Amazon IVS Console to generate a key pair, the private key is immediately downloaded to your machine and is . not saved in the console or available for further download Creating a Key Pair Offline If you'd prefer to create your own key pair, you can do that offline and then import the public key into the Amazon IVS console. Here's an example of creating a key pair with OpenSSL. $ openssl ecparam -name secp384r1 -genkey -noout -out demo-private-channel-private-key-1.pem
$ openssl ec -in demo-private-channel-private-key-1.pem  -pubout -out demo-private-channel-public-key-1.pem To import the public key, click on on the list page. Import Playback key Give the imported key a name, and browse for the key file (or paste the key file contents). public public This can also be done via the AWS CLI: $ aws ivs \
    import-playback-key-pair \
    --name demo-private-channel-playback-key-1 \
    --public-key-material "`cat /path/to/your/public-key.pem`" Or via the AWS SDK ( for JavaScript example). docs Private Channel Playback Now that we have created a private channel and a playback key pair, we are ready to generate signed playback authorization tokens. In your application, your business logic will dictate which users should be granted an authorization token. As mentioned above, maybe your application logic will first check that the user is a valid subscriber to the channel by querying a database. Or, in the case of geo restrictions, maybe you'll have to look up the IP address of the viewer to make sure it is associate with a region that is able to view the stream. The exact implementation will vary, but once you determine whether the viewer is indeed authorized to view the stream, you can proceed with the token generation method described below. Token generation will differ based on the language you're using on the backend. Refer to the for details on the requirements. Heads Up: user guide To generate a token with Node.js, we can use the library ( ). This library will simplify the process and allow us to sign our payload with the signing algorithm as required by Amazon IVS. Our payload must contain the , an to limit playback to a specific domain (or for all domains), and a token expiration timestamp. To sign the payload, we'll format it as below and use the private key that we created and imported above. jsonwebtoken docs SHA-384 Channel ARN access-control-origin * The token is used to start playback on a private channel, so the expiration period only applies to the time between token generation and playback start. It does not relate to the amount of time the user can play a given stream. Once playback is authorized, it can continue until the user disconnects or reloads the web page. Note: Here's one way to generate a token on the backend (via a microservice or serverless application) using Node.js. import jwt from "jsonwebtoken";

const getToken = () => {
  const privateChannelArn = process.env.DEMO_PRIVATE_CHANNEL_ARN; // private channel ARN
  const privateChannelPrivateKey = process.env.DEMO_PRIVATE_CHANNEL_PRIVATE_KEY; // playback private key

  const payload = {
    "aws:channel-arn": privateChannelArn,
    "aws:access-control-allow-origin": "*",
    "exp": Date.now() + (60 * 1000), // expires in 1 minute
  };

  const token = jwt.sign(payload, privateChannelPrivateKey, { algorithm: 'ES384' });
  return token;
} In our frontend, we can retrieve the token from the backend and append it to the private channel's . In the example below, I've trunacted the Playback Url ( ) for brevity. Playback Url streamUrl const streamUrl = `https://[truncated].m3u8?token=${token}`
const ivsPlayer = IVSPlayer.create();
ivsPlayer.attachHTMLVideoElement(document.getElementById('video-player'));
ivsPlayer.load(streamUrl);
ivsPlayer.play(); Summary In this post we learned how to create private channels, create and import playback keypairs, and generate a JWT token for playback authorization. To learn more about private channels, refer to the . user guide

Using Amazon IVS to Create a Private Channel for Authorized Live Stream Playback

Uso de Amazon IVS para crear un canal privado para la reproducción de transmisión en vivo autorizada

Utilisation d'Amazon IVS pour créer un canal privé pour la lecture de flux en direct autorisée

अधिकृत लाइव स्ट्रीम प्लेबैक के लिए एक निजी चैनल बनाने के लिए Amazon IVS का उपयोग करना

Amazon IVS を使用して、承認されたライブ ストリーム再生用のプライベート チャネルを作成する

Usando o Amazon IVS para criar um canal privado para reprodução de transmissão ao vivo autorizada

Sử dụng Amazon IVS để tạo kênh riêng để phát lại luồng trực tiếp được ủy quyền

使用 Amazon IVS 创建私有频道进行授权直播流媒体播放

Using Firebase Authentication with the Latest Next.js Features

সর্বশেষ Next.js বৈশিষ্ট্য সহ Firebase প্রমাণীকরণ ব্যবহার করা

Verwenden der Firebase-Authentifizierung mit den neuesten Next.js-Funktionen

Uso de la autenticación de Firebase con las últimas funciones de Next.js

Utilisation de l'authentification Firebase avec les dernières fonctionnalités Next.js

नवीनतम Next.js सुविधाओं के साथ फायरबेस प्रमाणीकरण का उपयोग करना

Usando o Firebase Authentication com os recursos Next.js mais recentes

Использование аутентификации Firebase с новейшими функциями Next.js

Firebase Kimlik Doğrulamasını En Son Next.js Özellikleriyle Kullanma

Sử dụng Xác thực Firebase với các tính năng Next.js mới nhất

将 Firebase 身份验证与最新的 Next.js 功能结合使用

This paper is available on arxiv under CC 4.0 license. Authors: (1) Ayei E. Ibor, Trustworthy Digital Infrastructure for Identity Systems, The Alan Turing Institute, United Kingdom; (2) Mark Hooper, Trustworthy Digital Infrastructure for Identity Systems, The Alan Turing Institute, United Kingdom; (3) Carsten Maple, Trustworthy Digital Infrastructure for Identity Systems, The Alan Turing Institute, United Kingdom; (4) Gregory Epiphaniou, WMG, University of Warwick, United Kingdom. Table of Links Abstract & Introduction Background Review of Literature Methodology Discussion of Findings Conclusion and Future Work & References 4. Methodology To underscore the importance of interoperability in developing countries, an investigation and comparison of current interoperability solutions in the identity ecosystem was performed. This investigation identified the current limitations of existing solutions and provided the basis for our findings, which are relevant to achieving interoperability for foundational identity systems in developing nations. As a guide, we have presented an approach for the investigation of requirements in Figure 2. The approach in Figure 2 is an easy stepwise guide. The steps include: i) Identify key processes against which baseline requirements should be derived: these processes are already identified in this work. ii) Define common interoperability criteria for eID schemes in candidate countries: this is an important step that must be fulfilled based on the identified key processes of step 1. iii) Baseline requirements against common interoperability criteria, select and compare country-specific eID schemes with current interoperability solutions such as eIDAS and X-Road: this step allows for the identification of common functionalities among eID schemes relative to defined common interoperability criteria. iv) Decide on the benchmarking requirements: it is important to decide on the benchmarking requirements to adequately scope the functional requirements useful for a fully functional interoperable system. v) Use benchmarking requirements against specific eID schemes: using benchmarking requirements against specific eID schemes is useful for identifying missing requirements and functionalities. vi) Identify and scope common functional requirements between eID schemes: this is a significant step that ensures that all requirements for interoperability are adequately captured. vii) Perform requirements gap analysis by identifying missing requirements between eID schemes for all processes defined and recording the differences between eID schemes: the missing requirements must be identified in this step concerning the processes defined in step 1. All significant differences between eID schemes must be recorded (e.g. in Africa and Europe). viii) Define missing capabilities to fulfil newly established requirements: it may be necessary to define missing capabilities in the investigated eID schemes that are relevant to the identified requirements to realise an interoperable identity system. However, this may not be the case where such eID schemes fulfil most of the baseline requirements for interoperability. 4.1 Current Interoperability Solutions For the creation of effective, long-lasting, and usable identity ecosystems, interoperability is essential. Two major interoperability solutions in the identity ecosystem are discussed in this research. These include the Estonian X-Road and the electronic identification and trust services (eIDAS) of the European Union. X-Road is a data exchange layer solution that implements the interoperability of information systems. It allows organisations to exchange data through a secure channel over the Internet. This solution is centrally managed but supports a distributed exchange layer that provides a standardised and secure channel for producing and consuming services (X-Road, 2023a, 2023b). X-Road enhances the confidentiality, integrity, and interoperability between organisations that rely on it for data exchange (Bakhtina et al., 2022; Saputro et al., 2020). Jointly implemented by Estonia and Finland, it has been adopted by several countries and organisations (Saputro et al., 2020). As stated in Solvak et al. (2019), X-Road establishes online connections between service providers and data registries (such as the Population Register, Health Insurance Register, etc.). The citizens who use the X-Road system only give their information to the government once, and the public authority then stores and exchanges the information among itself via the X-Road system (Bhattarai et al., 2019). By signing the messages with the X-Road member's signature key and using a mutually authorised Transport Layer Security (TLS) channel, security servers guarantee the integrity and secrecy of the exchanged messages. By recording the exchanged messages and routinely timestamping the message logs, the signed communications' long-term evidential value is protected. To obtain information on the validity of certificates and timestamp-signed messages, the security servers communicate with trust services. In terms of message exchange, the trust service calls are asynchronous (X-Road, 2023c, 2023d). To exchange messages between a service client and a service provider, three protocols are used. These include X-Road message protocol, X-Road message transport protocol, and OCSP response retrieval protocol (X-Road, 2020). When an interested party such as an organisation joins an X-Road ecosystem, certificates issued by a reputable Certification Authority (CA) are used to verify the identification of each organization and security server. Each security server serves as the technical entry point that manages access control on the organisation level during the data exchange process between registered X-Road members (European Commission, 2023a; X-Road, 2023a). The combination of timestamping and a digital signature ensures that data delivered via X-Road cannot be disputed. On the other hand, eIDAS is a European Union’s framework to ensure that electronic transactions between businesses, citizens, and public agencies are safer and more efficient, regardless of the European country they take place in. It is established by a European Regulation that was implemented in 2014. By introducing a common framework for eID and trust services, the eIDAS regulation makes it easier to supply business services across the EU. It encourages interoperability across the 28 EU nations, making certain that nations mutually recognise each other's electronic identities and trust services across borders (Mocanu et al., 2019). The goal of electronic identification is to completely transform how customers engage with online services. The Member States of the EU may choose to identify citizens electronically. A small number of Member States have created national programs to provide their citizens with electronic identity (eID), with greatly diverse architectures (Lips, Bharosa and Draheim, 2020). National systems, therefore, vary not only in the volume of citizen data they process but also in the degree of data protection they provide to this data. As claimed by Cuijpers and Schroers (2014) and Hölbl, Kežmah and Kompara (2023), businesses and customers can more easily access services or conduct commercial transactions by using electronic identity, or eID, to identify who they are (identification process) and demonstrate that they are who they claim to be (authentication process). Similarly, the regulation stipulates that it will be necessary for all EU nations to accept notified eID systems from other nations by September 2018. When conducting electronic transactions, especially those between firms and clients who are based in another EU country, trust services attempt to improve the trust of EU residents and enterprises (Sharif et al., 2022). The trust services in eIDAS as discussed in ANSSI (2023) and European Commission (2023b) include Electronic Signature (eSignature), Electronic Seal (eSeal), Electronic Timestamp (eTimestamp), Website Authentication Certificates (WACs), and Electronic Registered Delivery Service (eDelivery). eIDAS has been exploited by both the public and private sectors in the EU. Some of the sectors that have benefited from the regulation include the financial services, online retail, transport, and professional services sectors. One of the largest potential beneficiaries of eID and trust services is the financial services industry due to the possibility of enormous commercial opportunities and enhanced cross-border services (Cuijpers and Schroers, 2014; European Commission, 2023b). To meet rising client demand for online services as well as stricter compliance requirements, the identification, authentication, and safeguarding of transactions in the financial services sector are becoming increasingly digitised. 4.2 Comparison of X-Road and eIDAS as Interoperability Solutions The comparison of X-Road and eIDAS is given in Figure 3. This comparison is performed based on identity creation and management, proof of identity, creating and issuing identity credentials, issuing a derived credential, managing the identity credential lifecycle, granting access to an identity holder, data exchange mechanism, security of identification data, and privacy of identification data. This paper is available on under CC 4.0 license. Arxiv

The Methods We Used to Investigate and Compare Cross-Border Interoperable Identity Systems

ALTORISE

altorise.com

11-50 employees

Since 2014

COMPANY RANKING

ALTORISE

EVERGREEN INDEX #718

Representing this company? Click here to claim and customize this page!

Altorise WIKI

Apply to get your Tech Company News Page live on HackerNoon.com today!

ALTORISE

altorise.com

11-50 employees

Since 2014

COMPANY RANKING

ALTORISE

EVERGREEN INDEX #718

RELATED COMPANIESRANK

STORIES ABOUT ALTORISE

Representing this company? Click here to claim and customize this page!

Altorise WIKI

Apply to get your Tech Company News Page live on HackerNoon.com today!