Tension between the press and artificial intelligence (AI) giants is rising again. In recent weeks, several headlines and publishers have confirmed that they have blocked the robot that OpenAI runs on websites to suck up their content and train its AI models, including the famous ChatGPT conversational agent.
In France, Radio France explained that it had taken this step to avoid seeing "its content plundered without authorization." "This summer, we blocked the OpenAI robot, which was drawing on our content without our consent," explained its president, Sibyle Veil, at a press conference on Monday, August 28. Another public group, France Médias Monde (France 24, RFI) has done the same, as has TF1 for MyTF1 and TF1info.fr and the Sipa-Ouest-France group for its local information platform Actu.fr.
In the US, the New York Times also blocked the bots feeding OpenAI's models, according to The Verge. According to The Guardian, the Times was followed by CNN, Reuters, the Chicago Tribune, the ABC and Australian Community Media (ACM) brands such as The Canberra Times and The Newcastle Herald. Since then, CNN has completed the list of media with the Disney, Condé Nast, Hearst and Vox Media groups, the Bloomberg agency, and the Washington Post, The Atlantic, Axios, Insider and ESPN.
More broadly, these blockages reflect the desire expressed by many media outlets over the past few months to be paid for the use of press content by AI giants. Software such as ChatGPT or Bard (Google) are trained on articles produced by the media and are then able to create texts in response to requests from internet users on topics like current affairs... The threat is deemed all the greater as these agents will be integrated into search engines such as Google or Bing, which bring traffic to press sites.
The principle of respecting copyright and payment is widely defended in France by publishers' unions Geste and the Alliance de la presse d'information générale (General Information Press Alliance), as well as by media outlets such as Le Figaro and Le Monde, which has made contact with AI software manufacturers. AFP, for its part, has co-signed an article denouncing AI as a "threat to the financial viability" of the media. All of these – as well as the Financial Times, News Corp, New York Times, Springer and The Guardian – hope to negotiate "licensing agreements" for payment for AI training content.
OpenAI has already signed agreements with the AP agency and the Shutterstock image bank, but discussions with the media are likely to be tense and lengthy. Questioned in July, Google said it was premature to talk about licensing. In talks with OpenAI for several months, the New York Times would not rule out legal action against the creator of ChatGPT to obtain compensation, according to NPR radio. "We need fair value for the use of our content, past and future," said CEO Meredith Kopit Levien at the Cannes Lions Festival in June.