In recent years, the adoption of generative AI has profoundly transformed the way companies manage intellectual property. In particular, the integration of technologies based on Natural Language Processing and Large Language Models has opened up new possibilities, but also new questions. The topic of AI and, more broadly, intellectual property is now central for companies that want to innovate without exposing themselves to risks.
On the one hand, AI enables the analysis of enormous amounts of data; on the other, it raises complex issues related to ownership, data usage, and the protection of inventions, as well as the possibility of using AI systems for reliable analysis in the IP context.
Discover our AI software for the automatic extraction of information from technical documents.
Copyright and Generative AI: Who Really Owns the Content?
One of the most debated legal issues related to AI concerns the relationship between copyright and generative AI. When content is produced by an algorithm, determining who holds the rights is not straightforward. The issue of copyright in AI-generated content is closely linked to liability: who is legally responsible for AI-generated content? The user, the developer, or the platform provider? In many jurisdictions, this question still lacks a definitive answer, creating a grey area that can expose companies and professionals to significant legal risks.
Generative AI and Copyright: The Ghibli Style Case
A landmark case that contributed to igniting the debate on generative AI and copyright involves the spread of images created “in the style of” Hayao Miyazaki and Studio Ghibli. For many, it was one of the first moments when the power of AI appeared both fascinating and problematic: a simple prompt was enough to generate scenes that clearly echoed the poetic and recognizable aesthetics of the famous Japanese films. Although these were not direct copies, the results raised profound questions about copyright, generative AI, and plagiarism, especially regarding the boundary between inspiration and creative exploitation. This episode made it evident how even stylistic imitation can have legal and reputational implications, generating issues related to plagiarism or copyright infringement, and bringing intellectual property protection in the age of artificial intelligence to the forefront.
Use of Protected Data for Training AI and Liability
Another crucial issue concerns the use of protected data to train AI systems. Generative models learn from datasets often composed of both public and private information, but it is not always clear whether such data has been used in compliance with regulations. The topic of AI datasets and intellectual property infringement is therefore increasingly relevant. Companies must carefully monitor data provenance and implement control policies to avoid violations that could have legal and reputational consequences.
Patent Infringement with AI and Legal Protection
In the industrial context, one often underestimated risk is the potential infringement of existing patents in situations where analysis is conducted primarily through generative AI tools. Generative systems can in fact “hallucinate” and suggest technical solutions already covered by existing patents, without being able to distinguish between what is free and what is protected. Without proper oversight, the use of AI can lead to the development of non-patentable solutions or, worse, to unintentional infringement of third-party rights.
AI as Inventor: The DABUS Case
A particularly relevant case in the debate on AI and intellectual property is the DABUS case, an artificial intelligence system designed to autonomously generate inventions. Its creator, Stephen Thaler, launched a global strategy by filing patent applications in several countries naming the AI as the inventor. This initiative raised a crucial question: can an artificial intelligence system be legally recognized as an inventor? International responses have been far from uniform. In most jurisdictions, the applications were rejected on the grounds that the inventor must be a natural person. However, some exceptions emerged: in South Africa, a patent was granted listing AI as the inventor (although without a substantive examination), while in Australia a first judicial decision opened the possibility of recognizing an AI system as an inventor, emphasizing the importance of not hindering innovation. Despite these openings, the prevailing view remains anchored in a traditional model in which only humans can be recognized as inventors. The DABUS case, however, has helped stimulate global discussion among legislators, patent offices, and companies, highlighting an increasingly evident regulatory gap. In a context where AI plays a growing role in technological development, there is a clear need to rethink existing legal categories to avoid leaving innovations generated—at least in part—by autonomous systems without protection.
But When Analyzing Patents, What Results Does AI Currently Produce?
AI and Intellectual Property
Patents are an extraordinary source of technological and competitive knowledge. Analyzing them means understanding innovation trends and anticipating competitors’ moves. Generative AI can facilitate this process, but it is not sufficient on its own. The complexity of technical language and the structure of patent documents require advanced tools and rigorous methodological approaches.
Despite their capabilities, Large Language Models have significant limitations when applied to intellectual property. They do not truly understand technical concepts but rely on statistical correlations. This can lead to errors, inaccuracies, and so-called “hallucinations,” which are particularly risky in legal domains where reliability is essential: decisions based on incorrect information can have significant economic and strategic impacts.
Hybrid Approach and Legal Protection in Generative AI
To address these challenges, Erre Quadro proposes a hybrid approach that combines AI and human expertise, based on models trained on proprietary, continuously updated datasets integrated with expert oversight. This reduces errors, improves analytical quality, and ensures more reliable results. Human intervention becomes essential for interpreting data, validating information, and making strategic decisions.
Opportunities of Generative AI in the IP World
In intellectual property, we must now come to terms with AI: while it currently does not appear possible for AI to generate intellectual property, and the legal risks related to training data and copyright cannot be ignored, it is certainly a powerful tool for managing information. The key lies in developing a structured approach that integrates technological tools and human expertise. Only in this way can generative AI be transformed from a potential source of risk into a strategic lever for innovation.
Request a demo of our AI tools.
