{"id":113261,"date":"2025-10-31T01:39:19","date_gmt":"2025-10-31T01:39:19","guid":{"rendered":"https:\/\/www.newsbeep.com\/ie\/113261\/"},"modified":"2025-10-31T01:39:19","modified_gmt":"2025-10-31T01:39:19","slug":"hundreds-of-thousands-of-videos-from-news-publishers-like-the-new-york-times-and-vox-were-used-to-train-ai-models","status":"publish","type":"post","link":"https:\/\/www.newsbeep.com\/ie\/113261\/","title":{"rendered":"Hundreds of thousands of videos from news publishers like The New York Times and Vox were used to train AI models"},"content":{"rendered":"

Last month, The Atlantic dropped the latest investigation in its ongoing series on generative AI training data sets<\/a>. Staff writer Alex Reisner found that at least 15 million YouTube videos<\/a> had been used for training data by major technology companies, either for research or, in some cases, to build AI video products.<\/p>\n

The Atlantic\u2019s reporting focused over a dozen prominent training data sets that were either compiled or used by companies including Microsoft, Meta, Snap, Tencent, Runway AI, and ByteDance. The investigation shows how the unauthorized use of YouTube videos has been an essential contributor to the AI industry\u2019s recent leap forward<\/a> in AI video generation quality.<\/p>\n

\u201cMuch as ChatGPT couldn\u2019t write like Shakespeare without first \u2018reading\u2019 Shakespeare, a video generator couldn\u2019t construct a fake newscast without \u2018watching\u2019 tons of recorded broadcasts,\u201d writes Reisner.<\/p>\n

The Atlantic\u2019s story briefly mentions that more than 30,000 videos from the BBC were among the training data, alongside other YouTube channels focused on news. Using a searchable database published by The Atlantic<\/a>, I wanted to better understand the scale at which news channels had been targeted. In the same data sets, I found hundreds of thousands of videos that were taken from some of the most popular news publishers and news creators on YouTube, including The New York Times, The Washington Post, The Guardian, Al Jazeera, and The Wall Street Journal.<\/p>\n

\"The<\/a>For example, more than 88,000 videos were included from Fox News\u2019 YouTube channels, including its flagship account<\/a> and Fox Business<\/a>. Another roughly 70,000 videos were taken from the channels of ABC News<\/a> and its morning show, Good Morning America<\/a>. I also found more than 55,000 videos from Bloomberg\u2019s YouTube channels, including Bloomberg Originals<\/a>, Bloomberg Television<\/a>, and Bloomberg Technology<\/a>.<\/p>\n

Searching through Vox Media-owned YouTube channels in the database, I found more than 30,000 videos including explainers from Vox<\/a>, travel docs from Eater<\/a>, and animal tearjerkers from The Dodo<\/a>. Roughly 13,900 of those videos were from The Verge\u2019s official YouTube channel<\/a>, including iOS gadget guides<\/a>, episodes of its flagship podcast<\/a> The Vergecast, and interviews with Silicon Valley CEOs<\/a> like Mark Zuckerberg.<\/p>\n

YouTube CEO Neal Mohan has previously said that it\u2019s against the platform\u2019s terms of service<\/a> for other companies to download videos and use them for training data.<\/p>\n

\u201cIn order to survive, AI platforms know they need (and their consumers want) quality, credible content like ours that give their products relevance and purpose,\u201d said Lauren Starke, a spokesperson for Vox Media. \u201cThey\u2019re spending at unprecedented levels on AI infrastructure: chips, servers, and data centers that power their models. Yet when it comes to the content that makes those models useful \u2014 journalism, creative work \u2014 they\u2019ve comparatively spent next to nothing.\u201d<\/p>\n

In May 2024, Vox Media signed a partnership with OpenAI<\/a> for an undisclosed sum allowing the company to use its content for products like ChatGPT. Starke said Vox Media will continue to explore partnerships with AI companies that respect their work, but \u201cpursue legal remedies to protect our intellectual property, when necessary.\u201d<\/p>\n

\u201cWithout our quality content, the reality for these platforms will be: garbage in, garbage out,\u201d she said.<\/p>\n

\nThe Atlantic\u2019s database includes over a dozen distinct video training data sets, all of which have been used prominently in generative AI research and development. Some of those data sets have explicit links to commercial video generation models on the market.<\/p>\n

For example, I found 11,604 videos from the official YouTube channel of The New York Times across 11 different data sets in the database. Over 8,000 of these videos, though, came from a single training data source \u2014 Runway Gen-3<\/a>. Compiled by Runway AI, a company that has received backing from Salesforce, Google, and Nvidia<\/a>, this data set was made to train its flagship video generation model. When Gen-3 was released back in June 2024, it received rave reviews and was compared<\/a> to earlier iterations of OpenAI\u2019s Sora and Google\u2019s Veo models. Earlier this year, Runway AI was valued at $3 billion<\/a>.<\/p>\n

Among the thousands of videos from The New York Times in Runway Gen-3, there is a documentary<\/a> on JFK\u2019s assasination, a visual investigation<\/a> on the Hong Kong pro-democracy protests, a sit-down interview<\/a> with Barack Obama, and an opinion column<\/a> about Russian influence operations. An additional 382 videos are taken from the NYT Cooking YouTube channel, including viral recipes<\/a>, how-to baking guides<\/a>, and shortform street food docs<\/a>. (One caveat is that Runway AI may have omitted certain videos when it ultimately trained Gen-3.)<\/p>\n

An internal Runway AI spreadsheet<\/a> published by 404 Media<\/a> last year gives some insight into why YouTube videos from news publishers were targeted. The spreadsheet, called \u201cVideo sourcing \u2013 Jupiter\u201d (referencing the codename for Gen-3), lists thousands of channels that were marked by the company as high quality.<\/p>\n

In the document<\/a>, 27,000 videos from The Wall Street Journal\u2019s YouTube channel were tagged with: \u201clot of graphics, walkthroughs, \u2018show and tell.\u2019\u201d From CNET, 22,000 videos were described as \u201ctech reviews\u201d and tagged with the keyword phrase \u201cusing a laptop.\u201d From the Washington Post, 21,000 videos were labeled as \u201clots of newscaseter[sic] but plenty of b-roll.\u201d Another 35,000 videos from Good Morning America were tagged \u201cgargling,\u201d AI jargon for when a model superficially mimics something from its training without deeper \u201cunderstanding.\u201d<\/p>\n

From The New York Times\u2019 official YouTube channel, videos were listed with the description \u201cnyt video, op docs, b roll, talking, human subjects.\u201d Hundreds of NYT Cooking videos were tagged with the keyword \u201cscrambling eggs.\u201d This language gives some indication of the visual vernaculars \u2014 or even specific actions \u2014 Gen-3 was being trained to mimic.<\/p>\n

Since the model\u2019s release, major Hollywood studios have started folding Runway AI\u2019s products into their film, TV, and marketing production pipelines. According to a report from Bloomberg<\/a> this summer, Netflix is already using Runway AI tools in its \u201ccontent production<\/a>\u201d and Walt Disney Co. has similarly been testing out its software.<\/p>\n

\"\"<\/a>Meanwhile, there have been no reported licensing deals between Runway AI and the many news publishers whose work was included in the data set, including The Washington Post, Vox Media, the BBC, and The New York Times. Runway AI did not respond to a request for comment.<\/p>\n

\u201cThe Times has not authorized the use of videos that it publishes on YouTube for AI training purposes by any third party. As the owner of these works, the Times has the exclusive legal right to decide how and where our content is used \u2014 and are monitoring this closely,\u201d said a spokesperson for the Times, which is currently suing OpenAI and Microsoft<\/a> for allegedly using its text articles to train ChatGPT without permission. \u201cWe will continue to actively investigate infringement of our valuable intellectual property, and will enforce our rights as appropriate.\u201d<\/p>\n

Not all the training data sets in The Atlantic database have such clear ties to commercial AI video products. Some were used by the research arms of major AI companies, including Meta, Snap, Tencent, and Bytedance. This usage is public because employees disclosed it themselves in research articles.<\/p>\n

For example, a training data set called HD-VILA-100M was first collected by Microsoft Research Asia<\/a>, the company\u2019s research lab headquartered in Beijing, China. The Atlantic reported<\/a> that HD-VILA-100M was made available for download by Microsoft researchers and then used by a host of major AI companies in their own research and development.<\/p>\n

Meta used the data set to develop its text-to-video system \u201cMake-A-Video,\u201d<\/a> which was released back in 2022. A research lab at Tencent, the Chinese tech giant, used HD-VILA-100M to create a publicly-available dataset<\/a> that could rival the training data used by OpenAI for its Sora video generation model. Byetdance, the owner of TikTok, used the model to train its experimental text-to-video model MagicVideo<\/a>. Snap, the owner of Snapchat, used the model for research<\/a> into improving AI video captioning, video search tools, and text-to-video generation<\/a>.<\/p>\n

\"\"<\/a>Within HD-VILA-100M, as it was passed across the AI industry over several years, were thousands of YouTube videos owned by news publishers. That includes more than 13,000 videos downloaded from Fox News YouTube channels, roughly 6,300 from various DW channels, and another 5,520 from the Al Jazeera English channel, among others.<\/p>\n

While research using HD-VILA-100M has advanced video generation technology at each respective company, it\u2019s harder to draw straight lines between its usage and any one proprietary model or feature.<\/p>\n

Similarly, YT-Temporal-180M<\/a> is a dataset compiled by researchers at the University of Washington and the Allen Institute for AI, a nonprofit research organization. The Atlantic reported<\/a> that the data set is hosted on Google Cloud servers and available for download<\/a> through Hugging Face, a platform for sharing data sets and machine learning models. YT-Temporal-180M includes about 36,000 videos from Fox News, about 34,000 videos from Bloomberg, and roughly 31,000 videos from ABC News, among others.<\/p>\n

Since it first became available in 2021, YT-Temporal-180M has been downloaded from Hugging Face more than 1,450 times<\/a>. Many of the data sets identified and audited by The Atlantic remain available for download on Hugging Face to use freely for model training.<\/p>\n

\nMajor publishers were not the only news-focused YouTube channels I found. Videos from news creators \u2014\u00a0 independent channels that host news aggregation, talk shows, interviews, and political punditry \u2014 were scattered throughout the training data sets and sometimes rivaled the numbers from traditional news media.
\nI found several of the most popular progressive news channels on YouTube in the training data sets, including over 15,000 videos from The David Pakman Show<\/a>, a talk show that has more than three million YouTube subscribers. His videos were included in both HD-VILA-100M and YT-Temporal-180M, among others. Pakman, the founder and host of the program, confirmed he had not received any requests to use these videos for AI training.<\/p>\n

\u201cI understand that AI training often involves scraping large amounts of publicly available data, and that\u2019s part of how these systems improve,\u201d Pakman told me. \u201cWhen the use is this concentrated \u2014 a.k.a tens of thousands of videos from one creator \u2014 it feels less like incidental inclusion and more like large-scale extraction of intellectual property without consent.\u201d<\/p>\n

Wired previously reported<\/a> on how subtitles from Pakman\u2019s videos were used to train language models.<\/p>\n

Over 11,000 videos from The Majority Report with Sam Seder<\/a>, which has nearly 2 million subscribers on YouTube, were also in the data sets. When I spoke to Seder, he speculated that his channel offers AI companies a \u201cvisual and linguistic vernacular\u201d that\u2019s fundamentally different from mainstream news publishers. Those thousands of videos from The Majority Report include recorded livestreams, listener call-in shows<\/a>, and reaction videos<\/a>, all of which round up to a radio-jockey-style brand of political commentary<\/a>.<\/p>\n

Notably, very few of the most prominent U.S. conservative political commentators on YouTube were in the data sets. For example, there were no videos from Steven Crowder or The Rubin Report. There were 460 videos from Ben Shapiro\u2019s YouTube channel, which has over 7 million subscribers.<\/p>\n

Under YouTube\u2019s rules, when a creator uploads an original video, they automatically retain copyright<\/a>. That said, YouTube does have a carve out to use the content for its own AI training purposes. Earlier this year, CNBC reported<\/a> that YouTube had used a subset of videos on its platform to train Google\u2019s Gemini and Veo 3 models. This allowance does not extend to third parties.<\/p>\n

\"\"<\/a>News publications and news creators don\u2019t need to register their YouTube videos with the U.S. Copyright Office (USCO) to have a valid copyright claim. That said, registering videos by submitting an application and paying a filing fee does come with legal benefits, like the ability to sue for copyright infringement.<\/p>\n

The New York Times told me that it \u201cregisters its print edition and website on an ongoing basis with the US Copyright Office, including all underlying content.\u201d In many cases, YouTube videos from the Times that are based on print or web articles that have already been registered with USCO could be considered \u201cderivative works<\/a>\u201d and covered by the same filings.<\/p>\n

\u201cTaking content from creators like the Times without permission violates the law and will severely harm the market for original, independent reporting, which will diminish the ability of people to tell important stories, leaving the public less informed,\u201d a spokesperson for the Times told me. \u201cThe Times believes that the future success of this technology should not come at the expense of journalistic institutions.\u201d<\/p>\n

Seder, meanwhile, said none of the videos on The Majority Report channel \u2014 often five uploads per day \u2014 are registered with the USCO. As he puts it, he simply doesn\u2019t \u201chave the pockets\u201d to cover filing fees and retain legal counsel, especially when up against some of the largest companies in the world.<\/p>\n

He is comfortable with other creators pulling clips from his videos without permission, to a degree. After all, reaction videos are fuel for news creators across YouTube.<\/p>\n

\u201cPeople are using my content all the time, but they\u2019re adding commentary to it, and it is part of a conversation, and it is transparent \u2014 that\u2019s part of the ecosystem,\u201d said Seder. He sees the mass downloading of his channel for AI training in another light. \u201cWhat these [AI companies] are doing is fundamentally different. There\u2019s no reciprocity; it\u2019s only exploitative.\u201d<\/p>\n

Graphic of YouTube logo made up of dollar bills by Destrosvet<\/a> and used under Adobe Stock license.<\/p>\n","protected":false},"excerpt":{"rendered":"Last month, The Atlantic dropped the latest investigation in its ongoing series on generative AI training data sets.…\n","protected":false},"author":2,"featured_media":113262,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[20],"tags":[220,218,219,61,60,80],"class_list":{"0":"post-113261","1":"post","2":"type-post","3":"status-publish","4":"format-standard","5":"has-post-thumbnail","7":"category-artificial-intelligence","8":"tag-ai","9":"tag-artificial-intelligence","10":"tag-artificialintelligence","11":"tag-ie","12":"tag-ireland","13":"tag-technology"},"_links":{"self":[{"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/posts\/113261","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/comments?post=113261"}],"version-history":[{"count":0,"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/posts\/113261\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/media\/113262"}],"wp:attachment":[{"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/media?parent=113261"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/categories?post=113261"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/tags?post=113261"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}