Getty Images accused Stability AI, the company behind the text-to-image generative AI systems Stable Diffusion and DreamStudio, of using over 12 million of its copyrighted photos without permission to train their AI models. The case is one of many to emerge that highlights the tension between AI and the traditional understandings of ownership and authorship that our intellectual property rights are based on. Here, Dan Adams, founder of technology law firm Arbor Law, explains why there are no experts in this emerging landscape.
The pace of technological breakthroughs from artificial intelligence (AI) is outstripping the ability of legislators to update our intellectual property (IP) rights in the face of this new world. Those concerned about the IP implications of AI technologies might ordinarily turn to legal experts to advise them. However, we should be weary of anyone claiming expertise in this emerging legal space. As things currently stand, there are no experts here.
It is often the case that there is a lag between the introduction of a new technology and the development of a regulatory framework capable of addressing the challenges and opportunities that arise from it. In many respects, the recent emergence of AI, which is making global headlines on an almost daily basis following the launch of ChatGPT, is no different.
However, the challenge on this occasion is perhaps greater as the pace of technological change threatens to outstrip anything we have seen before. There is a recognition that some of the more archaic legislation will need adjusting to avoid stifling technological innovation, but weakening IP rights too much could mean that creators and inventors are not rewarded for their contributions and the creation of original work is disincentivised.
Early indications
Different countries will strike the balance in different ways. In the UK, prime minister Rishi Sunak (at the time of writing) has lent his support for unleashing technological innovation and making the UK a centre for AI development, a rhetoric that hints toward weakening traditional IP protection. However, there have been no changes to date.
The key development so far has been a government consultation, which laid out some of the challenges and defined the issues at stake. In addition to addressing the balancing act described above, the consultation highlighted how traditional approaches to defining authorship and ownership are challenged by the emergence of AI-generated works. If an AI system creates a piece of music or a new invention, who owns the rights? Is it the developer of the AI, the user, or the AI itself? These questions are yet to be definitively answered.
Another key issue, highlighted by the Getty Images case at the outset of this article, is whether the training of AI algorithms raises IP issues. AI systems rely on vast amounts of data to learn and improve. This data often includes copyrighted material, raising questions about whether using such data for training purposes constitutes a breach of IP rights. The issue of fair use and the rights of data owners need to be carefully considered to avoid potential legal pitfalls.
MORE FROM ARTIFICIAL INTELLIGENCE
The consultation raised the profile of these issues and will lay the foundation for future legislative changes, but for the foreseeable future, lawyers will have to work within existing regulations which were mostly designed for a pre-digital age. With an election campaign to take place first, and the strong possibility of a new government, it is unlikely we are going to see major changes by the end of this year, even while the pace of AI development continues unabated.
Although there remains the necessity to work within existing regulations, this does not rule out the possibility of judicial activism leading to unanticipated outcomes. As current copyright laws will inevitably be applied to scenarios that they were not designed for, courts may be tempted to apply their own interpretation to reverse-engineer existing laws and provide what they regard as the ‘‘right’’ outcome.
Implications for engineering
It appears likely that the first frontier will be the creative industries. In terms of the potential to challenge existing IP laws, it is here that recent breakthroughs in AI will have the more immediate impact in upending existing copyright laws. This will inevitably lead to the establishment of new legal precedents, so engineers might have the benefit of watching developments play out in this field first.
Engineering applications of AI tools like large language models (LLMs) will take more time to develop. Looking at current capabilities, AI tools can convert text into creative outputs like stories, lyrics or even pictures that replicate particular styles. For engineering applications, the most obvious being text-to-design, the existing tools are less well-suited to delivering workable outputs. Text to design is more complex than text to images and given the safety critical nature of many engineering designs, it will take longer to develop.
The issues around training appear different. Many of the most-high profile AI tools are trained on vast quantities of open-source material. ChatGPT, for example, is trained on the entire internet. Although this has already given rise to legal challenges, developing AI tools for engineering or manufacturing applications will require access to extensive data that is closely guarded intellectual property and not freely accessible on the internet. This will likely require carefully negotiated partnerships between engineering firms and those developing algorithms to provide the correct protection of the training data and govern the future uses of the any tools developed.
We can expect to see more high-profile legal cases and government consultations in the coming months, but it might be a while before concrete legislative changes are enacted. In this uncertain regulatory environment, engineers and their lawyers will have to operate within a regulatory environment that was primarily designed for a pre-digital age.
Dan Adams, founder of Arbor Law
Promoted content: Does social media work for engineers – and how can you make it work for you?
So in addition to doing their own job, engineers are expected to do the marketing department´s work for them as well? Sorry, wait a minute, I know the...