News publishers have raised concerns about A.I. chatbots like ChatGPT relying on copyrighted articles to power their technology. These publishers argue that developers of these tools disproportionately use news content.
The News Media Alliance, a trade group representing over 2,200 publishers including The New York Times, released research indicating that developers favor articles over generic online content to train the technology. They found that chatbots reproduce sections of copyrighted articles in their responses, leading the group to argue that A.I. companies violate copyright laws.
According to Danielle Coffey, the president and CEO of the News Media Alliance, this exacerbates an existing problem where tech companies like Google do not adequately compensate news organizations for displaying their work online.
Representatives for Google and OpenAI, the maker of ChatGPT, have not yet responded to requests for comment.
Generative artificial intelligence, the technology behind chatbots like ChatGPT, gained popularity with its ability to answer questions and complete tasks using information sourced from the internet and other sources. Other tech companies have also released their own versions of chatbots.
As the specific data fed into these models is often undisclosed, the News Media Alliance compared publicly available data sets used to train prominent large language models (which power A.I. chatbots) with an open-source data set of generic content scraped from the web.
Their analysis revealed that the curated data sets rely on news content significantly more than the generic data set, with a ratio ranging from five to 100 times. Ms. Coffey argued that this demonstrated a preference for quality content by those building A.I. models.
The report also identified instances where the models reproduced language used in news articles, indicating that publishers’ content was being retained and used by chatbots. As a result, the chatbots’ output competes with news articles.
“It genuinely acts as a substitution for our very work,” said Ms. Coffey, adding that their articles were being taken and regurgitated verbatim.
The News Media Alliance has shared their report’s findings with the U.S. Copyright Office’s study on A.I. and copyright law.
According to Ms. Coffey, this demonstrates a strong case the Alliance would have in court.
She also mentioned that the News Media Alliance is actively exploring the collective licensing of content from its members, who include major news and magazine publishers in the country.
Media executives have expressed various concerns about A.I., including its use of articles to train language models. They fear that chatbots could become the primary search tool, leading to reduced traffic from search engines to news sites. Additionally, many media workers worry about potential job replacements by A.I.