Getty Images sued AI company Stability AI in the UK, claiming that its image-generator tool, Stable Diffusion, infringed IP rights owned by Getty in the development, training, and use of the image‑generation AI model.
Judge Joanna Smith recognized the trademark infringement claim, specifically concerning Getty watermarks generated by Stable Diffusion users. Still, she dismissed the secondary copyright infringement claim, concluding that Stable Diffusion did not store or reproduce copyrighted works in a way that would amount to importation into the UK.
It should be noted that Getty did not pursue the primary claims related to unauthorized training, outputs, and databases due to a lack of evidence that the training occurred in the UK. Therefore, the judgment does not decide whether UK‑based web scraping and unauthorized AI model training would infringe copyright.
Catriona MacLeod Stevenson, General Counsel & Deputy CEO of the Publishers Association, said: “This is not the precedential AI and copyright moment that this case may, at one stage, have looked set to deliver. The judgment is hugely limited in scope, fell down on insufficient evidence rather than fundamental legal principle, and does not rule at all on the critical question of whether the training, development and operation of an LLM amounts to primary infringement of copyright in the UK (…) What the judgment does do is unmask the crippling impact even well-resourced companies like Getty face in protecting copyright in the absence of transparency requirements. It has had to spend millions tackling a single developer and will now have to continue to do so in another territory.”
Later in November, a German court delivered a landmark decision on copyright and AI training.
The case was brought by GEMA, the German collecting society representing around 100,000 music creators, over the use of lyrics by well-known German authors. The court ruled that OpenAI infringed GEMA members’ copyrights both by using the lyrics for training and by reproducing them in responses to users. Claims for injunctive relief, disclosure, and damages were upheld.
Although OpenAI argued that its models do not store or copy training data, and claimed that training and output production were in any event covered by the exceptions set out in articles 3 and 4 of Directive (EU) 2019/790, the Munich court dismissed these arguments in their entirety. Instead, the court found that the presence of the lyrics in ChatGPT’s parameters amounted to fixation, and that their reproduction in ChatGPT outputs represented additional infringing acts of reproduction and communication to the public.
Peter Kraus vom Cleff, Managing Director of the Börsenverein des Deutschen Buchhandels, stated: “The Börsenverein welcomes the decision of the Regional Court, which for the first time in Europe holds a major AI provider liable for the unauthorized use of copyrighted works. We can only hope that the previous defenselessness of creatives and cultural workers with regard to the unasked and unpaid exploitation of their services will come to an end through this and future court rulings. Without the mass use of third-party intellectual property, AI platforms would not have been able to open up billion-dollar markets. Therefore, the acquisition of licenses for training with copyrighted content is a requirement of appropriateness and fairness.”
Mikaela Zabrodsky, CEO of the Swedish Publishers’ Association, also celebrated the decision: “This is a historic verdict. For the first time in Europe, an AI company is held responsible for the unauthorized use of copyrighted works. We welcome the ruling, which confirms that copyright must also be respected in the development of AI services.”
In other sector news related to artificial intelligence, the European and International Booksellers Federation (EIBF) published its Charter on Artificial Intelligence, which it believes will help the publishing sector to maintain its sustainability and resilience. The 11 Guiding Principles proposed are:
- Copyright is the bedrock of our industry
- Copyright enables innovation
- Governance and accountability are needed to level the playing field
- A human-centric and ethical GenAI is the way forward for our industry
- Consumers have the right to know
- Environmental sustainability is key
- Competitiveness means that SMEs and micro-companies are not left behind
- Bookshops require compliant solutions
- Generative AI should be based on consent
- AI has the potential to support and help booksellers in their daily work
- We need research and assessments of the risks and impact of GenAI
Finally, IPA continues its advocacy work against the introduction of TDM/AI exceptions. For instance, IPA recently supported its Australian member (APA) in opposing the introduction of a Text and Data Mining (TDM) exception into Australian copyright law, achieving a successful outcome (read more here). Following a similar approach in Vietnam, IPA has argued that no TDM or AI exception should be enacted, whether through an AI bill or via IP legislation.