http://ipkitten.blogspot.com/2025/02/dutch-court-holds-that-tdm-opt-out-must.html

In a recent decision (ECLI:NL:RBAMS:2024:6563) on the application of the EU’s text and data (TDM) mining (TDM) exception in respect of RSS feeds, which may also be relevant to AI, the Amsterdam District Court (the Court) held that an opt-out from the TDM exception pursuant to the the Dutch transposition of Article 4 of the DSM Directive needs to be done in machine-readable means.  

Background

The Claimants, DPG Media, Mediahuis Netherlands and Mediahuis NRC (together, the Claimants) are all European news media companies. Between them, the Claimants publish a number of national and regional newspapers (including, for example, De Telegraaf, NRC, Volkrant, and Sport Niuws), and operate various websites associated with such newspapers. 
The Defendant, Knowledge Exchange BV, which uses the trading name “HowardsHome” (HowardsHome), is a commercial news aggregator which has offered, since 2001, a news alert service under the name “Howardshome nieuws” to approximately 60 clients, including public institutions, private companies, and publishers. It gathers information for its alerts mainly from publicly available internet RSS feeds of news articles. 
An RSS feed essentially sets out some basic data about each article, in a standardised format, including title, date/time of publication and a short excerpt, together with a hyperlink to the full article and any associated images. 
By February 2020, all of the Claimants had granted licences in their content to LexisNexis for the corporate market. In March 2020, HowardsHome also entered into an agreement with LexisNexis, whereby it would be allowed to purchase access links, headlines, snippets and full text of the licensed content of the Claimants from LexisNexis and integrate them into its own alert messages.
However, in March 2022, LexisNexis unilaterally terminated this agreement, alleging that HowardsHome had violated its terms and infringed the rights in the Claimants’ content by providing its clients with content sourced from either publicly available RSS feeds or scraped directly from the Claimants’ websites, some of which was behind a paywall.  
In a letter dated June 23, 2022, the Claimants informed HowardsHome that its large-scale automated searches of their news titles infringed on the Claimants’ right to make their content publicly available. The letter stated that HowardsHome’s services required the Claimants’ permission, which could be obtained from LexisNexis or another licensing company. 
HowardsHome denied any infringement on a number of bases, but the most important for present purposes (focusing on copyright and AI) were: 
  1. That the Claimants’ press publications were not protected by copyright;  
  2. That a temporary copying exception applied; and 
  3. That a TDM exception applied.
The Court’s ruling

Copyright in press publications

The Court found, in respect of the Claimants’ allegations of infringement of their press publishers’ right (introduced in Article 7b of the Dutch Neighbouring Rights Act by virtue of Article 15 of the DSM Directive), that these claims failed because HowardsHome had only used snippets of 150 characters (approx. 20 words) in its alerts, and the Claimants had failed to show that this could harm the normal exploitation of their news content (as per Article 15(1) and Recital 58 of the DSM Directive). 
HowardsHome’s primary defence in respect of the copyright infringement claim was, therefore, that it would be undesirable (and unlikely) for press publications that fall outside the scope of the press publisher’s right, to still be protected by copyright. 
The Court rejected this argument, noting that the two rights are not interdependent. Further, the Court found that, given that HowardsHome had used all of the Claimants’ RSS feeds without assessing whether or not each feed met the Infopaq (C-5/08) test for originality or not, the Court was entitled to presume for its current analysis that at least some of such feeds did meet the originality test and are therefore protected by copyright.  
Temporary copies
The Court then moved on to HowardsHome’s second defence, namely that its service involved the creation of temporary copies, where part of the RSS feed is copied and temporarily stored in a database or on its own servers, merely for temporary reading. The argument was that Article 13a of the Dutch Copyright Act (Auteurswet), which implements Article 5(1) of EU Directive 2001/29/EC (the InfoSoc Directive) would apply. 
The Court went through the five criteria set out by the Court of Justice in Infopaq II (C-302/10). These include, importantly, the requirement that the reproduction must lack independent economic value – i.e. it does not generate profit beyond what would be derived from legitimate use of the protected work. 
In this respect, the Court found that HowardsHome’s reproductions are central to its service, as they are sent to clients as alerts. They therefore held independent economic value and did not fall within the scope of Article 13a Auterswet. 
TDM 
Finally, the Court addressed HowardsHome’s argument that it could rely on the TDM exception in Article 15o Auterswet (implementing Article 4 of the DSM Directive). 
Since the Claimants had been unable to prove that HowardsHome’s alerts included or provided access to any information which was behind a paywall on the Claimants’ websites, the Court proceeded on the assumption that any information had been lawfully accessed in principle. 
The question was therefore whether the Claimants had effectively opted out of the TDM exception, and the Court expressly quoted from recital 18 of the DSM Directive in this respect, which states: “For content that is available online to the public, the reservation of rights should only be considered appropriate when implemented using machine-readable means, including metadata and the terms of a website or service. Reserving rights for text and data mining should not apply to other forms of use.
Crucially, the Court found that the Claimants had failed to prove that any TDM on their websites was explicitly reserved in machine-readable means, because the prohibition on automated searches which had been included in their evidence only excluded specific AI bots such as GPTBot, ChatGPT-User, CCBOT, and anthropic-ai. 
The Court found that this did not constitute a rights reservation “in machine-readable means in an appropriate manner“, and HowardsHome’s reliance on the TDM exception was therefore successful.
Three-Step Test

The Claimants nevertheless alleged that the three-step test derived from the InfoSoc Directive precluded HowardsHome from relying on any exception. By way of reminder, the three-step test states that exceptions to copyright infringement can only apply: 
  1. in certain special cases; 
  2. where they do not conflict with the normal exploitation of the work; and
  3. where they do not unreasonably prejudice the legitimate interests of the rightsholder.
AI-generated image of a Kat doing the “three step”

The Court focused on the second and third steps, which it considered together. It found that the core of the Claimants’ business model involved selling subscriptions, selling issues on paper and digitally, and selling individual articles upon request. The Claimants had failed, however, to provide any evidence that quoting content, which otherwise remains a paywall, affects their business. 

As for the alert service, which was provided by LexisNexis (pursuant to a licence from the Claimants), the Court found that while the product was not the same as that offered by HowardsHome, clearly the two companies were competitors. Nevertheless, the possibility that LexisNexis could lose revenue did not imply that the Claimants’ legitimate interests were unreasonably prejudiced. 
Based on the above, the claim of copyright infringement could not be upheld. 
Comment
Avid IPKat readers will recall that we recently featured commentaries on the decision of the Hamburg District Court in the LAION case (here and here), where that court had suggested that a reservation of rights formulated in “natural language” should be regarded as “machine-readable”. 
Clearly there is an inconsistency of interpretation in the EU regarding what constitutes a sufficient opt-out for the purposes of Article 4 of the DSM Directive, which is one of the issues that the EU Commission will be hoping to resolve through guidance, including the Code of Practice which is currently being consulted upon. 
In this Kat’s opinion, if the requirement for a “machine-readable” rights reservation were to be interpreted and applied as it was by the Dutch court in this case, it would appear to be make it extremely difficult for rightsholders to exercise work-level opt-outs of TDM in respect of certain categories of content (including, particularly, press publications), as well as downstream copies of works, or works licensed to third parties and therefore displayed on domains which are not within the rightsholders’ direct control. 
Outside of the EU, it will be interesting to see how the UK Government will seek to address these issues following the current consultation process, in which it disclosed that the preferred option (at least, at the outset of the consultation) was to follow the EU’s model. 
 


Content reproduced from The IPKat as permitted under the Creative Commons Licence (UK).