Generative AI: legal and regulatory issues
The Artificial Intelligences Generative AI (generative AI) is gaining increasing popularity in artificial intelligence applications. This growth has been particularly evident since the launch of Chat GPT by OpenAI. In this article, we will explore the legal and regulatory issues related to generative AI.
Generative AI: Definition and Applications
Generative AI is a category of artificial intelligence that can create new content, such as text, images, video, audio, code, or synthetic data. This broad field includes predictive algorithms and those that use external suggestions ("prompts") to autonomously generate content. Early applications focused on automatic image and audio correction.
Commercial and Economic Implications
While generative AI may be a fun pastime for individuals, it has significant commercial implications. It can be used in industries such as information technology, marketing, the film industry, publishing, and more. It can replace human activities, including software development, and support a variety of intellectual professions. However, there is a risk of eroding the economic value associated with human ingenuity.
Artists' Concerns and Data Provenance
Artists have raised concerns about the widespread use of AI, suspecting that training data could be obtained through web scraping from copyrighted human content. This raises questions about data provenance and the protection of AI-generated works.
Regulation of AI in Europe
The European Union is attempting to regulate AI through the AI Act, similar to the GDPR for data protection. The primary focus is on risk and liability. However, questions regarding intellectual property rights also arise.
Training Dataset and Copyright
In the legal field, lawsuits are emerging regarding the alleged illicit origin of training data. For example, Getty Images filed a lawsuit against Stability AI when its software returned images with a Getty Images watermark indicating copyright. A legal solution is being sought to compensate authors and copyright holders.
Problems with Class Actions
Several class action lawsuits in the United States have brought charges against AI vendors for misusing copyrighted data. However, proving plagiarism in AI-produced works is complex, as AIs use a wide variety of sources to produce output.
Data Source and Improper Use
The origin of the data used to train AIs is a complex issue. While the use of public databases is legal, indiscriminate scraping can raise copyright issues. Furthermore, the use of data by AIs for generative purposes can be considered the creation of derivative works, potentially infringing copyright.
Limits of Justice in the Context of AI
Tracing the origin of data used by AI is extremely complex, as machine learning processes involve processing large amounts of data in opaque ways. This presents challenges for courts and copyright protection.
The Future of AI Regulation
The European AI Act aims to promote transparency in the use of generative AI. However, its practical implementation raises challenges, especially for companies using pre-trained AI. Justice will often have to rely on the courts' legal wisdom to address these complex issues.
Copyright Ownership between Man and Machine
In the context of the interaction between generative AI and copyright, a fundamental controversy emerges: who should own the copyright of AI-generated works? This question opens a complex discussion involving legal, ethical and philosophical aspects.
Stephen Thaler's Advocacy: AI as Copyright Owner
Stephen Thaler has been a proponent of the idea that AI should be considered copyright holders, or, more broadly, intellectual property holders. This perspective is supported by those who see unique and original qualities in AI output, or who believe that the generative dynamics are completely independent of human input.
In the case known as DABUS (Device for Autonomous Bootstrapping of Unified Sentience), Thaler attempted to file patents attributing the invention to an AI, an unsupervised neural network he himself programmed. However, most courts have ruled that intellectual property rights are a human prerogative, relying on existing laws or on the principle that attributing rights to a machine appears to contradict human legal principles.
The only exceptions were Australia and South Africa, but even in these cases, the issue was controversially discussed and often overturned later. Overall, the global trend has been to maintain copyright as a human prerogative.
The Case of Thaler II's Creativity Machine
Thaler recently sought recognition for his Creativity Machine as the sole author of a two-dimensional artwork titled "A Recent Entrance to Paradise." What makes this unique is that the Creativity Machine doesn't receive prompts from its creator but learns autonomously from provided datasets, completely eliminating human intervention.
However, a critique of this perspective might argue that, even if AI operates autonomously, humans are involved in data validation and the system's initial programming. In other words, for an AI to produce output, some form of human instruction must have been provided, even remotely.
Recently, the DC District Court ruled that, as with patents, copyright attribution is a human prerogative. This reinforces the idea that, despite the astonishing capabilities of AI, copyright should remain under human control.
The Role of Copyright
Copyrights serve several purposes, including the moral gratification of authors, the possibility of obtaining economic benefits, and the legal protection of assets of economic value against claims or injury by third parties. The idea of granting such rights to a machine, which lacks emotions, desires, or awareness, raises fundamental questions.
A comparison can be made with similar situations, for example, the use of software to create special effects in films or the generation of art by AIs like Dall-E. In these cases, despite the significant contribution of AI, the names of the software are not included in the film credits as "co-authors," and the artists using AI as a tool must still provide significant input to achieve the desired results.
Ultimately, the goal of attributing copyright to software seems inconsistent with our current legal and social system, which does not consider machines to be equal to human beings.
Protecting the Works Produced by AI
An important aspect to consider is the protection of AI-generated works. Many courts and legislations have addressed this issue differently. In the United States, for example, there is a rather conservative position that a priori excludes the registration of works created exclusively by AI. However, there are exceptions, such as the case of artist Kristina Kashtanova, who registered a comic book created with the help of AI but stated that she nevertheless made personal editorial and stylistic choices.
The United Kingdom, on the other hand, claims to be among the few countries to protect computer-generated works even in the absence of a human creator. However, it defines the author of such works as “the person by whom the necessary arrangements for the creation of the work were made”, underlining the importance of human contribution.
The European Union appears to take a more flexible approach, suggesting that when AI assists an author in the creative process, the existing copyright legal framework still applies. However, the definition of a work's originality is closely linked to human personality, judgment, and technical skill.
Conclusions
Issues surrounding copyright and artificial intelligence remain complex and evolving. The main trend seems to be to keep copyright human-centered and to exclude the attribution of rights to AI.
The definition of creativity and originality is being challenged by the presence of generative AI capable of producing astonishing results, yet most courts and legislation continue to require significant human input to ensure copyright protection.
The future may see greater discernment regarding the role of AI in artistic creation and the creative process, but at the moment,
