In an ongoing battle between Google and France’s competition authority regarding copyright protections for news snippets, the Autorité de la Concurrence has recently imposed a hefty €250 million fine on the tech giant. The Autorité de la Concurrence announced this decision on Wednesday, which, at the current exchange rate, amounts to approximately $270 million.
As per the competition watchdog, Google has failed to uphold certain commitments it made to news publishers in the past. However, the decision is particularly significant because it eliminates another modern aspect: Google’s utilization of news publishers’ content to train its generative AI model, Bard/Gemini.
The competition authority has raised concerns about Google’s failure to inform news publishers about the use of their copyrighted content by GenAI. This is in reference to previous commitments Google made to ensure it engages in fair payment discussions with publishers regarding the reuse of their content. 
Issues of copyright and competition
In 2019, the European Union implemented a pan-EU digital copyright reform that expanded copyright protections to include news headlines and snippets. News aggregators, like Google News, Discover, and the “Top Stories” feature box on search results pages, used to gather and showcase these news stories on their platforms without providing any financial compensation.
Google initially attempted to circumvent legal obligations by disabling Google News in France. However, the competition authority swiftly intervened, deeming their independent action an abuse of their dominant market position that could potentially harm publishers. The intervention essentially compelled Google to negotiate agreements with local publishers regarding the reuse of content. In 2021, Google faced a hefty $592 million fine due to significant breaches in its negotiations with local publishers and agencies, as determined by the competition authority.
The tech giant deemed the sanction “disproportionate” and expressed its intention to appeal. However, it later attempted to resolve the disagreement by making a series of commitments and retracting its appeal. The French Autorité has accepted the commitments, which involve the sharing of vital information with publishers and engaging in equitable negotiations.
Google has entered into copyright agreements with numerous publishers in France as part of its agreement with the Autorité. So businesses in this area are subject to strict regulations.
No interest
Google has agreed to accept the Autorité’s latest findings without contesting them in return for a streamlined process and a financial payment.
However, Sulina Connal, the managing director for news and publishing partnerships, expressed frustration in a lengthy blog post, stating that the authority’s fine does not adequately address the issues at hand.
The blog post indicates that Google is determined to put an end to the ongoing situation. Connal, the author, emphasizes the importance of moving forward and shifting the focus towards sustainable methods of connecting people with valuable content. Additionally, Google aims to establish positive relationships with French publishers.
Given the advent of generative AI and the intense race to release new tools, Google’s approach to the content reuse issue has undergone a significant shift.
Engaging in GenAI training within the framework
France’s competition authority’s action today emphasizes its focus on Google’s use of news publisher and agency content to train its AI foundation model and its associated AI chatbot service Bard (now known as Gemini).
In July 2023, Google introduced Bard, its generative AI tool, without informing the copyright holders or the Authority, according to a press release.
Google’s defense on this matter has two main aspects. In its blog post, Google asserts that Article 4 of the EU Copyright Directive already covers the use of web content to enhance newer products like generative AI.
Article 4 of the Copyright Directive outlines a provision for “exception or limitation for text and data mining,” which pertains to the reproduction and extraction of legally accessible works and other subject matter for the purpose of text and data mining.
However, according to the press release from the Autorité, it is still under investigation whether the exemption is applicable in this case. (It is important to highlight that the clause in question pertains to “lawfully accessible”works”—Google, being legally obligated to inform copyright holders about the use of their protected works, seemingly did not fulfill this obligation in this particular instance.)
“The question of whether using news content to train an artificial intelligence service falls under neighboring rights and protection remains unanswered,” stated the competition authority. “Nevertheless, the Autorité believes that Google has violated its commitment #1 by neglecting to notify publishers about the utilization of their content in the training of Bard.”
Google’s blog post briefly acknowledges the importance of the EU AI Act. Nevertheless, the legislation is still awaiting final adoption by the European Council before it can come into effect.
The upcoming AI legislation will also require developers to comply with the bloc’s copyright rules. Furthermore, it enforces transparency obligations by mandating the implementation of a policy that upholds EU copyright law. Additionally, it necessitates the provision of a comprehensive summary of the data utilized to train general-purpose AI models, like Gemini/Bard.
News publishers that have used protected content for GenAI training could potentially benefit in the future from the new requirement for model makers to publish a training data summary. This may help them receive fair remuneration under EU copyright law.
There is no way to avoid the technical aspects.
The Autorité also highlights that Google did not offer a technical solution, at least until September 28, 2023, for publishers and press agencies to choose whether their content can be used to train Bard without it impacting the display of their content on other Google services.
Publishers and news agencies previously had the option to prevent Google’s various services, such as Search, Discover, and Google News, from indexing their content. This required them to include a specific instruction to block the indexation. The negotiation specifically includes these services for revenue related to neighboring rights, underscoring the Autorité’s commitment to closely scrutinize Google’s opt-out processes in the future.
News publishers had the option to include a “noindex” tag in the robots.txt file from July to September 2023 to prevent Google’s AI model training from using their content. This robots.txt file is typically located in the root folder of web servers and serves as a guide for search engines, providing them with specific instructions. The web crawler at Google examines the instructions in those files to index websites.
However, using a “noindex” tag will result in your website being removed entirely from Google. In September 2023, Google introduced a more detailed approach by implementing a “Google-Extended” rule that differs from the existing “noindex” rule. By choosing not to participate in the Google-Extended instruction, web publishers express their lack of interest in contributing to the enhancement of Gemini’s existing and upcoming models.
Additional limitations
The Autorité is also penalizing Google for various other concerns regarding its negotiations with French news publishers. The Autorité has determined that Google did not adequately provide them with the necessary information to ensure a fair bargaining process for the remuneration of their content.
In its press release, it stated that Google’s information to publishers regarding its methodology for calculating payment amounts was “especially unclear.”
The report also emphasized that Google failed to meet the non-discrimination criteria, which aim to guarantee fair and equal treatment for publishers. It criticized Google’s decision to set a “minimum threshold” for payment, arguing that it introduces unfair discrimination among publishers. All publishers, regardless of their situation, receive zero remuneration below a certain threshold, according to its press release.
In addition, the Autorité criticized Google’s calculations of “indirect income,” stating that the proposed “package” did not align with previous decisions or the appeal judgment of the Court of Justice in October 2020.
Additionally, it was noted that Google did not follow through on its promise to revise compensation agreements in accordance with its commitments.
