China

The courts in China continue to issue ground-breaking decisions concerning Gen-AI and copyrights. While in November 2023 the Beijing Internet Court ruled that AI-generated images are works of authorship (see here), last month (February 2024) the Guangzhou Internet Court ruled that a service provider of a Gen-AI tool is liable for copyright infringement for allowing its users to generate substantially similar and derivative works of various copyrighted works. The copyrighted works in question were taken from the Ultra Series, a Japanese science fiction media franchise featuring the “Ultraman”, which the court regarded as famous and thus assumed that the defendant had access to them.

One of the interesting aspects of the Guangzhou Internet Court’s ruling was its interpretation of China’s Interim Measures for the Management of Generative Artificial Intelligence Services of August 2023, and finding that the Gen-AI service provider did not employ the requisite measures thereunder; most notably, that it did not implement mechanisms for infringement complaints and reports; that it did not identify the AI-generated output as such; and that it failed to address potential infringement risks in its Terms of Use.

UK

The lawsuit brought by Getty Images against Stability AI with the High Court of Justice in January 2023 has reached another milestone, with Stability AI filing its defense last month.

Getty alleges that Stability AI has been using images from Getty’s websites to train Stable Diffusion‘s text-to-image Gen-AI model without authorization. In its defense, Stability AI repeated the claims previously raised in attempts to have the case dismissed (which the High Court of Justice rejected), and argued, inter-alia, that no training took place in the UK (implying that the English courts have no jurisdiction over the matter), that the development of Stable Diffusion did not itself involve the reproduction of any copyrighted works of Getty in the UK; that it never authorized its users to infringe works owned by Getty, which undermines the claim of secondary infringement; and that any AI-generated image that resembles a Getty image should be seen as a pastiche (“an artistic work in a style that may imitate that of another work”), which is highly likely to constitute fair dealing.

The defenses raised by Stable Diffusion are representative of the cross-border legal barriers facing  copyright holders trying to enforce their rights vis-à-vis developers of Gen-AI tools, most notably fair dealing/use and secondary/contributory infringement.

US

In February 2024 the news outlets The Intercept, Raw Story, and Alternet all filed lawsuits against OpenAI for using their articles to train ChatGPT without authorization. An interesting argument raised by the news outlets concerns the fact that OpenAI created tools in late 2023 to allow copyright owners to block their work from being incorporated into training sets, which suggests that OpenAI had reason to believe that use of copyrighted material in its training sets may constitute copyright infringement.

Irrespective of such tools, OpenAI’s position in these lawsuits is likely to be similar to the one it adopted in the Tremblay v. OpenAI lawsuit, where it filed and won a motion to dismiss claims related to vicarious copyright infringement (see here).

New Service Offering

Herzog has been working with clients across various industries to address the latest challenges and changes brought by Gen-AI tools. We (Technology Regulation and IP teams) offer “AI and Law” workshops and Q&A sessions, and provide customized guidelines for clients that require assistance in navigating the evolving world of AI.

For more information >> click here