Recently, Getty Images filed a lawsuit against Stability AI, accusing it of using copyrighted images to train its AI models without Getty’s consent. The question may also arise whether popular generative AI tools like Midjourney use databases containing copyright protected material without the consent of the rights holders to generate their images.
Recent lawsuits, which include class actions, will help determine to what extent AI developers may use copyright protected works to train their models.
As mentioned in our previous article, the owners of copyrights may rely on their exclusive rights of reproduction and public communication to prevent AI developers from using their works. In this article, we will address how contract law can help in protecting intellectual property that is publicly available, in particular via websites.
The case law of the CJEU on web content use
Several years ago, the Court of Justice of the European Union (CJEU) addressed issues related to the protection of online content, particularly concerning “deep linking” and “framing”, and this case law may help to define the boundaries when using works for AI training purposes.
Deep linking is when a hyperlink directs the internet user to a specific, generally searchable web page, rather than to the website’s home page. Framing is when the internet user clicks on a hyperlink and sees the content of another website in a frame on the original website, without other information or advertising on the other website.
In the landmark Svensson ruling, the CJEU held that the provision on a website of clickable links to works freely available on another website does not constitute an act of communication to the public (CJEU 13 February 2014, C‑466/12). As a result, it is legitimate to offer links to other publicly available websites, where copyright protected works are being shown, without having to seek the prior consent of the copyright owners. But the CJEU also ruled that, where a clickable link makes it possible for internet users to circumvent restrictions put in place by the operator of the other website, the party offering those clickable links (called “deep links”) is communicating to a new public and is acting in breach of the other website operator’s copyrights. A link may not be used to circumvent access restrictions to a work that is no longer accessible or that is only accessible to a restricted audience (for instance, paying subscribers), otherwise the party offering such a link is liable for copyright infringement.
In the BestWater ruling, the CJEU found that embedding a freely available copyright protected work on another site via “framing” does not violate copyright law either, provided that the work is not being communicated to a new public and that the technical means of communication is not different from the original means of communication (CJEU 21 October 2014, C-348/13). The framing of works via a hyperlink is only allowed to the extent that those works are freely available on the original website. Only then is the communication to the public allowed, which means that if website operators contractually prohibit the use of their works, others may not use those works via deep linking or framing.
Contractual protection of web content
It has become common for website operators, especially in the media sector, to include contractual clauses explicitly prohibiting third-party use of their content. These clauses, often found in the website’s terms of use, state that the content cannot be copied, reproduced or used without consent. The enforceability of these clauses depends on explicit or implicit acceptance by the user. If users are required to accept the terms before having access to the content of a website (e.g. by ticking an “I accept” box near the full text of the terms), these clauses will be valid and binding. Website operators are encouraged to adopt explicit acceptance mechanisms of their terms of use to better protect their content from misuse.
Contractual protection via the terms of use is not only recommended for copyright protected works, such as articles, photographs and songs, but also for databases. Like the owners of copyrights, the producer of a database also has the right to prohibit the reproduction and public communication of their database. They have a sui generis right to prevent the reproduction and public communication of (at least) a substantial part of their database (cf. EU Directive 96/9/EC). AI companies wishing to train their models by using other parties’ databases must not only respect the sui generis rights of the producers, but must also respect the contractual restrictions that the producers may impose when making their databases available online.
This article was written by Marine Uytters, with the valued assistance of AI. It is part of the series “AI and intellectual property rights”, written by the IP lawyers at CMS in Belgium. All the articles are available on our website. All your feedback and questions are welcome at [email protected].
Social Media cookies collect information about you sharing information from our website via social media tools, or analytics to understand your browsing between social media tools or our Social Media campaigns and our own websites. We do this to optimise the mix of channels to provide you with our content. Details concerning the tools in use are in our Privacy Notice.