{"id":660828,"date":"2025-04-05T01:55:42","date_gmt":"2025-04-04T22:55:42","guid":{"rendered":"https:\/\/en.buradabiliyorum.com\/openais-models-memorized-copyrighted-content-new-study-suggests\/"},"modified":"2025-04-05T01:55:42","modified_gmt":"2025-04-04T22:55:42","slug":"openais-models-memorized-copyrighted-content-new-study-suggests","status":"publish","type":"post","link":"https:\/\/buradabiliyorum.com\/en\/openais-models-memorized-copyrighted-content-new-study-suggests\/","title":{"rendered":"OpenAI’s models ‘memorized’ copyrighted content, new study suggests"},"content":{"rendered":"
\n

A new study<\/a> app<\/a>ears to lend credence to allegations that OpenAI trained at least some of its AI models on copyrighted content.<\/p>\n

OpenAI is embroiled in suits brought by authors, programmers, and other rights-holders who accuse the company of using their works \u2014 books, codebases, and so on \u2014 to develop its models without permission. OpenAI has long claimed a fair use<\/a> defense, but the plaintiffs in these cases argue that there isn\u2019t a carve-out in U.S. copyright law for training data.<\/p>\n

The study, which was co-authored by researchers at the University of Washington, the University of Copenhagen, and Stanford, proposes a new method for identifying training data \u201cmemorized\u201d by models behind an API, like OpenAI\u2019s. <\/p>\n

Models are prediction engines. Trained on a lot of data, they learn patterns \u2014 that\u2019s how they\u2019re able to generate essays, photos, and more. Most of the outputs aren\u2019t verbatim copies of the training data, but owing to the way models \u201clearn,\u201d some inevitably are. Image models have been found to regurgitate screenshots from movies they were trained on<\/a>, while language models have been observed effectively plagiarizing news articles<\/a>.<\/p>\n

The study\u2019s method relies on words that the co-authors call \u201chigh-surprisal\u201d \u2014 that is, words that stand out as uncommon in the context of a larger body of work. For example, the word \u201cradar\u201d in the sentence \u201cJack and I sat perfectly still with the radar humming\u201d would be considered high-surprisal because it\u2019s statistically less likely than words such as \u201cengine\u201d or \u201cradio\u201d to appear before \u201chumming.\u201d <\/p>\n

The co-authors probed several OpenAI models, including GPT-4 and GPT-3.5, for signs of memorization by removing high-surprisal words from snippets of fiction books and New York Times pieces and having the models try to \u201cguess\u201d which words had been masked. If the models managed to guess correctly, it\u2019s likely they memorized the snippet during training, concluded the co-authors.<\/p>\n

\"OpenAI
An example of having a model \u201cguess\u201d a high-surprisal word.<\/span>Image Credits:<\/strong>OpenAI<\/span><\/figcaption><\/figure>\n

According to the results of the tests, GPT-4 showed signs of having memorized portions of popular fiction books, including books in a dataset containing samples of copyrighted ebooks called BookMIA. The results also suggested that the model memorized portions of New York Times articles, albeit at a comparatively lower rate. <\/p>\n

Abhilasha Ravichander, a doctoral student at the University of Washington and a co-author of the study, told TechCrunch that the findings shed light on the \u201ccontentious data\u201d models might have been trained on.<\/p>\n

\u201cIn order to have large language models that are trustworthy, we need to have models that we can probe and audit and examine scientifically,\u201d Ravichander said. \u201cOur work aims to provide a tool to probe large language models, but there is a real need for greater data transparency in the whole ecosystem.\u201d<\/p>\n

OpenAI has long advocated for\u00a0looser restrictions\u00a0on developing models using copyrighted data. While the company has certain content licensing deals in place and offers opt-out mechanisms that allow copyright owners to flag content they\u2019d prefer the company not use for training purposes, it has lobbied several governments to codify \u201cfair use\u201d rules around AI training approaches.<\/p>\n<\/div>\n

If you liked the article, do not forget to share it with your friends. Follow us on\u00a0Google News<\/a><\/span>\u00a0too, click on the star and choose us from your favorites.<\/span><\/strong><\/p><\/blockquote>\n

\n

If you want to read more like this article, you can visit our Technology<\/a><\/span> category.<\/strong><\/p>\n<\/blockquote>\n

Source<\/a><\/span><\/p>\n","protected":false},"excerpt":{"rendered":"

A new study appears to lend credence to allegations that OpenAI trained at least some of its AI models on copyrighted content. OpenAI is embroiled in suits brought by authors, programmers, and other rights-holders who accuse the company of using their works \u2014 books, codebases, and so on \u2014 to develop its models without permission….<\/p>\n","protected":false},"author":1,"featured_media":660829,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"fifu_image_url":"https:\/\/techcrunch.com\/wp-content\/uploads\/2025\/03\/GettyImages-1466243153.jpg?resize=1200,705","fifu_image_alt":"","footnotes":""},"categories":[18],"tags":[77337,76255,87244,141199],"class_list":["post-660828","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-technology","tag-ai","tag-study","tag-copyright","tag-openai"],"_links":{"self":[{"href":"https:\/\/buradabiliyorum.com\/en\/wp-json\/wp\/v2\/posts\/660828","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/buradabiliyorum.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/buradabiliyorum.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/buradabiliyorum.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/buradabiliyorum.com\/en\/wp-json\/wp\/v2\/comments?post=660828"}],"version-history":[{"count":0,"href":"https:\/\/buradabiliyorum.com\/en\/wp-json\/wp\/v2\/posts\/660828\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/buradabiliyorum.com\/en\/wp-json\/wp\/v2\/media\/660829"}],"wp:attachment":[{"href":"https:\/\/buradabiliyorum.com\/en\/wp-json\/wp\/v2\/media?parent=660828"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/buradabiliyorum.com\/en\/wp-json\/wp\/v2\/categories?post=660828"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/buradabiliyorum.com\/en\/wp-json\/wp\/v2\/tags?post=660828"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}