To produce quality journalism, THE “ New York Times » “invests a huge amount of time, money, expertise and talent”. But to produce their ChatGPT artificial intelligence model, what do OpenAI and Microsoft use? Many articles from New York Timesresponds the American media which has therefore decided to take legal action against the two companies.
The question of copyright within AI was already raised by artists, translators and writers, who felt they were being copied by generative artificial intelligence, so it is the turn of everyday life to fight the battle. This Wednesday, he launched proceedings in a federal court in New York against OpenAI, creator of the ChatGPT software, as well as Microsoft, its main investor, for copyright infringement.
Millions of articles
At issue, according to the complaint, is the development model for generative AI, and in particular ChatGPT, which “is based on massive learning models built by copying and using millions of copyrighted Times articles.” “.
“As the complaint explains, Microsoft and OpenAI used our work to develop and market their generative AI products without having permission from the Times,” a spokesperson for the daily told AFP, believing that it was This is indeed a “violation of copyright in terms of content and journalistic work”.
And the financial consequences are not negligible, estimates the daily, which considers that the damage suffered, in fact and from a legal point of view, represents “several billions of dollars”. Questioned by AFP, neither Microsoft nor OpenAI have yet made any comments.
A potential breach
The newspaper claims to have wanted to avoid getting into legal confrontation with Microsoft and OpenAI, by having sought to negotiate an agreement allowing it to “receive adequate compensation for the use of its content”.
OpenAI has already reached an agreement with the German media group Axel Springer, however the agreement does not concern the use of the group’s media content to train its generative AI but the integration of this content into the responses to requests from ChatGPT users.
In this case, the New York Times’ legal action could open a breach and represent a potential new source of revenue for media around the world.
Because generative AIs, whether texts or images produced in response to written queries, like ChatGPT or Midjourney, are built on an algorithm trained using the immense amount of data available online, notably via social networks, most often without the authorization of their creators. Enough to fuel some trials, as well as articles to report on them. The AI will have enough to feed itself for a while…