The decision by a Belgian court on 13 February 2007 holding Google liable for copyright infringement has attracted enormous interest. Google included titles and snippets of Belgian newspaper articles in its Google News service. It also provided links to cached copies of the articles in its general search results. Here we give a detailed account of the Belgian court's judgment and comment on its implications for search engines.
Google, Inc. originally launched its Google News service in 2002. In January 2006 it launched the service in Belgium. On 9 February 2006 Copiepresse, an association of Belgian newspaper publishers, petitioned the Brussels court for an order for 'saisie description'. The court granted this on 27 March 2006 and appointed an expert to investigate. This court order was served on Google on 27 April 2006. The expert submitted his report on 6 July 2006. By letter of 13 July 2006 counsel for Copiepresse formally requested Google to remove its members' newspaper articles from Google News and the Google cache. Google did not respond to this letter.
Copiepresse started substantive proceedings in Belgium on 3 August 2006 and obtained a default judgment against Google on 5 September 2006. Google applied to set aside the order, but it was confirmed by the summary judge on 22 September 2006. Google then made a further application to have the claim dismissed, resulting in the judgment under discussion.
The rival contentions
Google described Google News as a specialised search engine based on indexing press articles published on the internet. Copiepresse however, considered that the Google News service went beyond a simple search engine service and acted as a "portal to the written press".
Google's principal activity is in providing its well known general search engine, described in the judgment thus:
"Google stipulates that its search engine is made up of indexation software or "robots" ... that trawl through websites moving through page after page, at regular intervals and in an entirely automatic way in order to list these in an index of all the web pages accessible to the public and the corresponding web address for each (called the URL address); Internet users can consult the index by means of keywords entered in the search bar, the search engine then displays the reference lists of pages available including the keywords searched and proceeds to an automated classification by relevance."
Google's general search engine displays a blank screen and search box to the user. Google News displays on its home page, in addition to a search box, a series of article titles, each accompanied by a short extract from the article itself, automatically selected and organised by topic. The titles constitute hyperlinks to the original articles on the newspaper publishers' websites. In support of its argument that Google News is more than a simple search facility Copiepresse relied heavily on the fact that Google News presents these items to the user on the Google News home page before any search is carried out.
Google's general search results also include links (labelled 'Cached') enabling the user to call up the HTML text copy of an article stored for a period in Google's own caching servers. However, 'Cached' links are not provided in Google News.
Although Copiepresse did not object to the provision of hyperlinks to the newspaper publishers' sites, it did object to the following:
- The display on Google News of the titles and first few lines of the newspaper articles.
- The 'Cached' links into Google's cached version of the articles. Although not available from Google News, the 'Cached' link was provided as part of Google's general search facility on Google.be.
In response to Copiepresse's complaints about Google News, Google made a general point that the same material was in any event available to users searching though its general search service. The Google News service was based on indexation similar to that carried out for Google's general search service and was therefore a search engine, not a site distributing news. Google then argued several specific legal points:
- Google was not reproducing and communicating protected works to the public, the user being referred by a hyperlink to the publisher's site. Google therefore only made the article more accessible by means of the hyperlink. The court concluded that this point was irrelevant since Copiepresse's objection was based not on linking to the publishers' sites but on the reproduction of the titles and extracts from the press articles. It was unambiguously apparent that Google News reproduced and communicated to the public on the homepage of its site (these elements being accessible on the Google News site itself by the simple consultation of this site), the press article titles and extracts from some of these articles.
- The titles of the newspaper articles and the displayed first sentence or sentences of the articles were not original elements benefiting from the protection of copyright law. The court found that some of the titles appeared to be sufficiently original to qualify for copyright protection, while others did not. The articles themselves were copyright works, so the question was whether reproducing the opening lines could to amount to substantial reproduction. The court concluded that at least in some cases it would. Google therefore reproduced and communicated to the public works protected by copyright.
- That in any event Google could rely on the citation and news reporting exceptions to copyright infringement under Belgian copyright law.As to the citation exception, the court found that this did not apply since Google limits itself to listing the articles and classifying them in an automatic way. Google News did not carry out any analysis, comparison or critique of the articles, which were not commented on in any way. As to news reporting, the court commented that Google's reliance on this exception was contradictory to Google's position that it was only a search engine, not an information portal. There was no commentary on the news, only extracts from articles grouped by topic. The exception did not apply where reproduction of the protected works was the principal object rather than secondary to the reporting. Google could not therefore rely on the news reporting exception.
In response to the complaint about the Cached link in its general search results, Google argued that it only copies the HTML code of websites into its cache, containing text elements and links but no images. Google maintained that it is the user, not Google, who creates a copy of the work so that the user is the author of any reproduction or communication to the public, the only involvement of Google being the supply of the installations intended to make it possible or to realise a communication to the public by internet users.
The court rejected this analysis as inaccurate. Google stored copies of web pages in its memory and the fact that it only stored HTML code was irrelevant. There was a digital reproduction from the moment that the material was stored and Google was the author of that reproduction. Google made that reproduction available on its own website via the 'Cached' link. Unlike with hyperlinks that referred to the site of origin, by consulting the 'Cached' link the user was consulting the document on the Google website. Google's role was consequently not limited simply to providing facilities. Google had reproduced and made available to the public a copy of the original documents stored in its own memory.
Finally, Google argued in response to all Copiepresse's complaints that if it were found to make copies of works protected by copyright and/or to communicate them to the public, the newspaper publishers had explicitly or implicitly consented to that.
The point of general interest here concerns implied consent. Google argued that it was open to the publishers to use technical means (robots.txt files and metatags) to exclude their publications from being indexed by search engines. These were standard methods known worldwide. Insofar as the publishers did not use them they were explicitly or implicitly agreeing to have their pages indexed and accessible via the cached links.
The court rejected this argument. It held that copyright is not a right of opposition but a right for prior authorisation. This meant that the authorisation had to be obtained in a certain way, prior to the intended use. It could not be maintained that "the use on the sites of robot files implies a certain and explicit agreement of the site publishers to the use of works included on the site as concretely used by Google News (all the more as this service only arrived on the Belgian market in January 2006 or, it appears, at a time when these robot-files were already integrated)". In the same way an unconditional authorisation for referencing could not be deduced from the absence of technical protection. That appeared to be all the more the case here, as Google News did not merely reference the articles but reproduced the title and an extract.
Electronic Commerce Directive
The court also held that the hosting and caching provisions of the Electronic Commerce Directive did not assist Google. Neither protection applied since it was Google's own behaviour that was in question, not the content of sites to which it permitted access. Further, the caching protection did not apply because it was not the temporary storage required for indexation that was in dispute, but making it available by means of the 'Cached' link.
Google argued that the order against it should be restricted to the Belgian sites Google.be and Google.news.be. Copiepresse argued that it should not be so limited, as the articles were accessible on Google's French site (google.fr) and on google.com. The judge found that since Google had stated that that it was only possible to comply with the order by completely de-indexing the newspaper publishers' sites, and that would result in their being removed from all Google News sites worldwide, there was no point in limiting the order to Google's Belgian sites.
This decision, which Google has said it will appeal, illustrates the difficulties that face search engines as they develop and extend their activities into areas beyond general purpose search. These challenges are especially pronounced in European countries, which unlike the USA generally lack a flexible doctrine of fair use of copyright works. Instead the norm in Europe is a series of narrowly crafted exceptions which often do not read well on to technological advances.
The long-established position of search engines should mean that if there ever was a question of general search activities infringing copyright under European copyright laws, they ought by now to have acquired legitimacy by virtue of implied licence. Hence Google's keenness in this case to present Google News as merely a specialised application of its general search engine. However, as this case demonstrates, implied licence can be a perilous position from which to extend into new areas where it may be arguable that the activity is not merely search and the public utility of the activity may be less generally accepted.
The Belgian court considered the question of implied licence and found that the newspaper publishers' failure to use standard technical exclusion methods such as the Robots Exclusion Standard and metatags did not amount to an implied licence. This makes an interesting contrast with Field v Google, Inc. (U.S.D.C. Nevada, 19 January 2006), in which the court held that the plaintiff had impliedly licensed inclusion in Google's cache by not setting exclusion parameters in his site's metatags or robots.txt file, when he was familiar with those processes. In that case the judge paid particular attention to the steps taken by Google to educate users as to how to prevent Google's robot from indexing their site or providing a 'Cached' link, and the fact that Google removed the 'Cached' links to the plaintiff's site as soon as they heard about his lawsuit.
Another potential problem with relying on implied licence is that the licence can be withdrawn. That ought not to be an issue with a service such as Google News, since Google has publicly stated that it operates an opt-out policy ("... if a content owner asks us to remove his or her content from our web search results, we do. If a newspaper does not want to be part of Google News, we take the paper's stories out." http://googleblog.blogspot.com/2006/09/our-approach-to-content.html). In the Copiepresse case the issues are complicated by the confusion over Google's apparent lack of response to communications from Copiepress or the court. So even if the court had accepted the implied licence argument, on the basis of the facts found by the court the period for which it would have applied might have been relatively short.
This illustrates the importance, if an online intermediary is relying upon implied licence, of having good procedures to respond to licence withdrawals in whatever form they are received. Clearly it is most convenient for search engines if website owners use standard technical means to do this such as the Robots Exclusion Standard and metatags. However it is difficult to see how the beneficiary of a bare implied licence could validly stipulate that that is the only mechanism by which the licence may be revoked.
The complaint based on the 'Cached' link was, as can be seen from the account of the judgment above, separate from the complaint about Google News. The 'Cached' link was available on Google's general search results, but not on Google News. Again, the court reached a different conclusion from that in the U.S. Field v Google case. There the court found, among other reasons for holding in Google's favour, that Google's delivery of cached content to users when they clicked on the 'Cached' link was non-volitional and therefore not an act of direct infringement. The provision of the cached link was additionally found to be fair use and also protected under the Digital Millennium Copyright Act.
Lastly, the territorial aspect of the order is of interest. Like the French judge in LICRA v Yahoo! (the Nazi memorabilia case) the judge in this case appears to have been little concerned by the possible extra-territorial effects of his judgment either as a result of declining to limit his order to the Belgian sites, or due to the practical effect of his order on Google. It is reminiscent of one of the earliest cases involving internet content, when CompuServe in December 1995 removed several hundred newsgroups from its systems worldwide as a result of the actions of a prosecutor in Bavaria.