French authorities have fined Google €270 million for using news content to train its AI model, Bard, without properly informing publishers. The French Competition Authority (FCA) believes that Google did not negotiate fairly with publishers over the use of their content. Google, on the other hand, argues that its use of news content falls under the scope of the EU Copyright Directive. The incoming EU AI Act, however, is expected to bring more transparency to the way companies use data to train AI models.
The FCA investigation began in 2020 following complaints from French news publishers’ alliance, Alliance de la Presse d’Information Générale (APIG). APIG accused Google of leveraging its dominant market position to exploit news publishers without fair compensation. The FCA’s decision comes just weeks after Google announced a new licensing program to pay publishers for their content in France.
According to the FCA, Google failed to properly inform publishers about how their content was being used to train Bard. The FCA also found that Google did not give publishers a real choice over whether or not their content could be used. Google claims that it has always complied with European copyright laws and that its use of news content is fair and legal under the EU Copyright Directive.
The EU Copyright Directive allows news publishers to control the reproduction and communication of their work to the public. However, it also includes exceptions for certain purposes, such as quotation, criticism, and parody. Google argues that its use of news content to train Bard falls under the exception for quotation.
The FCA disagrees with Google’s interpretation of the EU Copyright Directive. The FCA believes that Google’s use of news content goes beyond the scope of the quotation exception. The FCA also argues that Google’s dominant market position gave it an unfair advantage in negotiations with publishers.
The FCA’s decision is a significant blow to Google. The €270 million fine is one of the largest ever imposed by the FCA. The decision could also have implications for other tech companies that use news content to train their AI models.
The incoming EU AI Act is expected to bring more transparency to the way companies use data to train AI models. The EU AI Act will require companies to disclose the source of their training data and to assess the potential risks of their AI models.
Google’s decision to use news content to train Bard without properly informing publishers highlights the need for more transparency in the development of AI models. The EU AI Act is a step in the right direction, but more needs to be done to ensure that AI models are developed and used in a responsible manner.
