
Artificial Intelligence (AI) is revolutionising various industries but it poses significant challenges for intellectual property law. A recent talk at UCL from Gadi Oron, Director General, CISAC highlighted critical issues regarding AI's interaction with IP, particularly around the use of original works in AI training and the complexities of copyright in AI-generated content.
AI Training and Text/Data Mining
One of the fundamental aspects of AI development involves training models using vast datasets, often incorporating original works. This process, known as text and data mining, involves reproducing content to create comprehensive datasets. The EU permits text and data mining unless the rights holder expressly reserves their rights (Art 4 DSM Directive). This creates a proactive need for rights holders to reserve their rights to prevent their works from being used in AI training.
Practical Challenges for Rights Holders
However, the practicalities of reserving rights are fraught with challenges. For instance, configuring a website to exclude AI chatbots like ChatGPT can be technically complex. Rights holders may also face the daunting task of having to constantly scan the market to understand new AI services and their methods of operation. This raises the question: Is it appropriate to shift the burden of protection to rights holders? Moreover, if an author does not have a website, how can they effectively opt out of text and data mining? There's also the issue of whether a current opt-out can affect data that has already been mined.
Copyright in AI-Generated Works
The copyright status of AI-generated works remains a contentious issue. Cases like the "Monkey Selfie" concluded that works created without human intervention do not qualify for copyright protection. But what happens when a work is a combination of AI and human input? Recent case law, including Thaler and a Czech ruling on Dall-E, suggests that merely providing a prompt is insufficient for copyright protection. The US Copyright Office's decision in the "Zarya of the Dawn" comic underscores that human elements like arrangement of images and independently created text can be protected, but not the AI-generated content itself.
What does the AI future look like for IP?
The crystal ball gazing offered no certainties. A couple of particularly observations included:
The economics of AI and IP: the risk that denying copyright protection to AI works could create an uneven playing field for human creators who charge royalties for their work whilst AI works are royalty free.
New tools: these are being created to help distinguish between human and AI contributions to creative works as well as the relevant creative works that were used. The technology is still in its infancy and not entirely reliable.
Curation is key: even for works completely created by AI there is a human element in choosing the final wording or image. The importance of this aspect of human creativity is likely to increase in proportion to the uptake of AI. Curation is probably most akin to the database test for originality (selection or arrangement of material in a database). Although in this context, the output may not necessarily be a database.
To find out more about the issues raised in this blog contact Rosie Burbidge, Intellectual Property Partner at Gunnercooke LLP in London - rosie.burbidge@gunnercooke.com