Last week, a German musicians’ association achieved a significant legal triumph over OpenAI. According to the court, the development of the GPT 4 and 4o models involved copyright violations, and some of the model’s outputs also infringed on copyrights. This victory strongly supports the argument made by those who view it as merely a machine for plagiarism.

Those who harbor dislike towards OpenAI are likely to find resonance in intellectual property law expert Andres Guadamuz’s recent assessment of the ruling from the University of Sussex. Guadamuz highlights that while the decision and its consequences are somewhat convoluted, it could ultimately be advantageous for copyright owners in the long run.

This suggests that prominent figures in copyright—such as music stars, actors, and best-selling authors—may now be realizing the potential monetary gains from this technology, even if smaller creators may not reap similar benefits.

GEMA, a German organization representing composers, lyricists, and publishers, took legal action against OpenAI on behalf of stakeholders connected to nine well-known German songs. This is akin to advocating for the rights of composers and lyricists behind nine American songs ranging from “Soak Up the Sun” by Sheryl Crow to “Happy” by Pharrell Williams.

Unlike obscure lyrics sourced from a local band’s website for training data, these songs are iconic cultural references that would have been repeatedly included in training data in various forms such as altered versions or parodies, fragments, excerpts, and snippets.

The lawsuit was based on ChatGPT being able to provide responses like “What is the second verse of [the German equivalent of “No Scrubs” by TLC]?” even after disabling its web browsing feature. The court ruled that while ChatGPT’s responses might be incomplete or inaccurate at times, they still constituted further copyright infringement. It clarified that individuals seeking lyrics through ChatGPT are not infringing copyrights; instead, it is OpenAI that bears responsibility. Furthermore, since ChatGPT outputs can be shared via links, OpenAI was found to be distributing infringing material without authorization.

OpenAI is now required to disclose how frequently the lyrics were used as training data, any profits generated from them, cease storing them, refrain from reproducing them again, and potentially face financial penalties later on.

In a recent UK case involving Getty Images and Stability AI, a different outcome was reached where an AI model like Stable Diffusion that does not store or replicate copyrighted works was deemed non-infringing. Guadamuz’s analysis delves into why this discrepancy occurred and how machine “memorization” played a pivotal role in the German court’s decision compared to the UK ruling.

Contrary to the Getty judgment, this new ruling aligns with prevailing intellectual property legal doctrines in the digital age—asserting that similar copyright rules should apply whether it’s a playable CD or a CD-ROM.

While Guadamuz raises concerns about how this ruling addresses “memorization,” suggesting an attempt to establish non-memorization as the legal standard using EU data-mining regulations contradicts existing laws. However, he anticipates a future shift towards establishing some form of licensing market amid evolving interpretations of this case by companies over time.

As with Sora 2’s approach towards copyright and likenesses which gained acceptance from actors and copyright holders over time, a framework is emerging aimed at revenue sharing (future AI-generated income) with owners of copyrighted content. The emergence of generative AI initially unsettled copyright holders worldwide by introducing new dimensions of perceived copyright violations. Nevertheless, key stakeholders are gradually embracing generative AI as they envision potential financial gains ahead. This trend is evident in major U.S. record labels partnering with formerly litigated entities like Udio.

However, concerns linger among lesser-known copyright stakeholders—such as emerging artists, writers, and creators—who fear their work may become marginalized or obsolete in this new content landscape. The benefits for these individuals remain uncertain amidst these transformative shifts.