A US court (in Delaware) recently handed down a precedent-setting ruling on a lawsuit filed by a copyright-owner, the media and technology conglomerate, Thomson Reuters, against the artificial intelligence platform, Ross Intelligence. The Delaware court ruled that the defendant had infringed the plaintiff’s copyrights, and that the defendant is not entitled to any of the statutory protections, including reliance on the “fair use” doctrine as a defense.
The lawsuit
Thomson Reuters filed its lawsuit in 2020, alleging that Ross Intelligence had copied some of Thomson Reuters’ copyrighted materials from the legal search engine, Westlaw (Westlaw’s expert lawyers add headnotes (legal abstracts) above the legal documents). The court ruled that the copied materials were indeed protected, and that none of the defense’s arguments held water, and therefore, rejected all of them.
The court’s rulings
First, the court ruled that Ross Intelligence’s use of copyrighted materials belonging to Thomson Reuters constitutes infringement of Thomson Reuters’ copyrights. The court held that, since Ross Intelligence’s training of its AI platform involves, inter alia, copying of protected materials belonging to Thomson Reuters, as well as other actions, these actions constitute copyright infringement.
Second, the court rejected Ross Intelligence’s defense argument that the use of copyrighted materials is sheltered under the fair use doctrine.
This issue is at the core of disputes in such proceedings, since AI platforms claim that, even if training of the AI system prima facie involves copyright infringement, they should be granted protection by virtue of the fair use doctrine, which releases them from responsibility for infringement on the grounds of appropriate legal policy, which is based on the aspiration to legitimize technological progress.
The examination of “fair use” is based on four parameters, and the court ruled in relation to each parameter as follows:
- The purpose and nature of the use: the court ruled that Ross Intelligence’s use of the Westlaw database was commercial and not transformative; i.e., it was not intended to serve any other purpose other than a commercial one, and that at issue is blatant copying of copyrighted materials.
- The nature of the copyrighted work that was used: the court ruled that, notwithstanding the fact that the legal articles contained facts (which per se are not protected), the copied articles contained numerous, highly creative analyses authored by editors employed by Thompson Reuters.
- The qualitative and quantitative volume of use, relative to the entire copyrighted work: the court ruled in favor of the defendant, Ross Intelligence, in relation to this parameter, because although it had copied a considered volume of materials, the platform’s output to its users was at a lower volume.
- The impact of the use on the value of the copyrighted work and its potential market: the court ruled in favor of Thompson Reuters in relation to this parameter. The court rule that Ross Intelligence had harmed the existing and potential market for AI-based search engines and therefore, caused significant harm to Thompson Reuters.
Another key consideration was the fact that, prior to the infringement, Ross Intelligence had contacted Thompson Reuters to receive its permission to use its copyrighted materials. Thompson Reuters refused and, despite this refusal, Ross Intelligence used Thompson Reuters’ materials, knowing full well that at issue are copyrighted materials and that use thereof, without Thompson Reuters’ consent, is unlawful.
The court reasoned that Ross Intelligence intended to compete against the Westlaw database by using content it took from the database, which was not authorized, while attempting to develop an alternative market for it using a competing search engine. Therefore, the court ruled that the fair use doctrine does not apply in this case.
Implications of the precedent-setting ruling
As stated, this is the first court ruling (which might be appealed) adjudicating fascinating legal issues raised by interactions between intellectual property law and advanced technologies, including artificial intelligence. AI technology systems need to train their core systems and do so largely using enormous databases of works that might be protected by copyrights, which immediately opens a Pandora’s box of legal complexities.
This US court ruling attests that the protection of copyrighted works remains strong, even in the age of AI systems, and sends a clear message that operators of AI platforms are not allowed to do whatever they feel like doing with copyrighted works. Nevertheless, it is important to recognize the fact that Ross Intelligence did not operate a generative AI platform (which creates content on its own, such as ChatGPT or Midjourney), and it could be that this influenced the court’s ruling.
It is important to note that quite a few similar legal proceedings are underway in the United States, as well as in China, Canada and the UK. This precedent-setting ruling undoubtedly deals a blow to AI platforms and, although it might only be a bruise, its implications could be far-reaching. The ruling in the Thomson Reuters case might serve as a precedent for lawsuits in other courts in the future. However, it is also possible that – at least in some of the proceedings – the courts might differentiate between their cases and the ruling in the Thomson Reuters case, inter alia, due to the fact that the defendant was not operating a generative AI platform, and therefore, they might reach different conclusions.
In any case, operators of AI platforms that train their systems using copyrighted materials need to be highly cognizant of the implications of this ruling, since the precedent set could expose them to significant legal liability towards the copyright owners whose works they copied.
The Israeli Ministry of Justice published a position statement partially addressing this issue at the end of 2022. The ministry’s view at that time was that the use of copyrighted materials for the purpose of machine training is sheltered, in most instances, under the statutory “permitted use” arrangements, and therefore, will not be responsible for copyright infringement. This position reflects the desire to not lag behind in the development of AI systems and to incentivize desirable competitive activities in this developing industry. The Israeli ministry’s position, as stated, is inconsistent with the American court’s ruling in the Thompson Reuters case. It is definitely possible that, to the extent that other courts in the United States and around the world hand down similar rulings, the Israeli Ministry of Justice will be forced to reconsider its position, in order to fall in line with the prevailing jurisprudence.
[View source.]
