Brazil: AI Authorship and the curious case of Bill 303/2024

Brazil: AI Authorship and the curious case of Bill 303/2024

The rapid advancement of artificial intelligence (AI) is transforming industries, including innovation and intellectual property (IP). As AI systems increasingly generate creative Works and inventions autonomously or with minimal human input, they challenge traditional IP laws.

At the heart of the debate is the question of authorship and inventorship for AI-generated creations. Traditional IP frameworks assume a human creator behind every invention or creative work. This human-centric approach raises complex issues when an AI system produces something new without direct human involvement. This paper explores Brazil´s legal landscape regarding the authorship of AI-created inventions and highlights legislative responses to this emerging challenge.

In Brazil, the main current legislative project on AI is Bill 2338/23, which regulates the use of Artificial Intelligence. Its goals are to establish foundational principles for AI development and use, promote innovation while ensuring ethical standards and create a regulatory environment that balances technological advancement with the protection of individual rights, including privacy and data protection.

The bill emphasizes transparency in AI algorithms, accountability for AI decisions, and encourages investment in AI research and development to enhance Brazil´s global competitiveness. As of November 30, 2021, the bill was sent to the Senate and is awaiting discussion and vote.

Specifically addressing AI and intellectual property, on February 20, 2024, a new bill was introduced in the Brazilian Congress aiming to recognize an AI system as an inventor. Bill (PL) 303/2024 proposes amending the Industrial Property Law (Law No. 9,279/1996) by adding to Article 6 the following provision: “In case of inventions autonomously generated by an artificial intelligence system, the patent may be applied for in the name of the artificial intelligence system that created the invention, which can be considered the inventor and holder of the rights inherent to the invention.”

The author of this proposal cited the AI DABUS case developed by Stephen Thaler, who filed patent applications in Brazil and several other countries listing the AI as the inventor. Although his applications were dismissed, they sparked global discussions about recognizing AI as an inventor.

The congressman behind the bill argues that amending the Industrial Property Law is necessary to “adapt Brazilian legislation to the reality of technological innovations, specifically regarding inventions autonomously generated by AI systems.”

He believes that recognizing AI systems as inventors and holders of patentable inventions can accelerate innovation, boost economic growth, and create industries and employment opportunities. However, the bill does not explain how an AI system would have legal personality under national legislation to acquire patent rights – a key issue highlighted by the reporting congressman assigned to the bill.

Although the establishment of an AI system as inventor is already debatable (as the human condition is presumably inherent to the development of industrial property rights), recognizing an IA system as holder of patent rights inherent to the invention is a step further.

It is important to note that the Brazilian Patent and Trademark Office (BPTO) expressly rejects the indication of an AI system as the inventor in patent applications filed in Brazil, based on guidance from Legal Opinion No. 24/2022, where this issue was directly analyzed.

In a report issued on October 7 by the assigned reporting congressman to the Bill, the original proposal faces significant criticisms and requires adjustments and amendments. First, the absence of legal personality in AI systems is a major obstacle. Current legislation and legal doctrine allow only natural or legal persons to hold patent rights; assigning rights to non-human entities would require a fundamental redefinition of legal personality, which is a major legal challenge not faced in the referred Bill.

Second, there is ambiguity in the ownership of rights. If an AI system is considered the inventor, questions arise about who owns the intellectual property rights: the AI developer, the system operator, or the entity providing data for training the AI? Additionally, attributing authorship to machines raises concerns about diluting the concept of human creativity and innovation, impacting the value placed on intellectual work.

Third, it is imperative to define new criteria to evaluate the inventiveness, novelty, and industrial applicability of AI-generated creations, considering they might follow different standards than those established for human inventions. These changes could influence not only the intellectual property system but also the innovation market as a whole by altering the balance between protecting human and AI-generated inventions.

In order to address these issues and balance AI’s impact on innovation, the report proposed the following changes:

1. even when an invention or utility model is developed with partial or full assistance from AI systems, patent ownership will be granted to the human author or an entity.
2. patent protection terms for inventions or utility models developed with AI assistance should be modified. Specifically, Article 40 of Law No. 9,279/1996 would establish that invention patents developed with predominant AI assistance will have a validity period of five years, while those developed entirely autonomously by AI will have a validity of three years from the filing date.

And lastly, for utility models, these terms would be reduced to three and one years, respectively, depending on the degree of AI participation in the inventive process. Article 35 should also be amended to include, among the technical requirements, the classification of the degree of AI assistance in the patent application.

Bill (PL) 303/2024 is being assigned to a new reporting congressman, as the initial one left the Science, Technology, and Innovation Commission.

Globally, the prevailing view is that only natural persons can be recognized as inventors under patent law. The DABUS cases have sparked debate but largely reinforced the need for human inventorship.

In 2019, the USPTO rejected patent applications listing DABUS as the inventor, stating only natural persons can be named. U.S. courts upheld this decision, affirming that an inventor must be human under current law. In March 2023, the U.S. Copyright Office clarified that works generated entirely by AI aren’t eligible for copyright protection unless a human author exercises creative control.

The European Patent Office (EPO) rejected DABUS-related applications, emphasizing that the inventor must be a human with legal capacity.

In 2023, the BPTO issued a final rejection decision on prosecution of the DABUS patent application, due to the impossibility of designating or naming an AI as the inventor in a patent application filed in Brazil, based on the provisions of said Art. 6 of the IP Law, as well as the provisions of the Paris Convention and the TRIPS Agreement. In Brazil, the inventor of a patent application must be capable of being an individual having rights, possessing legal personality.

The German Federal Court of Justice (FCJ) ruled (June 11, 2024) that the natural person who significantly influenced AI must be identified as the inventor. This reinforced that while AI-generated inventions can be patented, inventorship is reserved for humans.

In a very recent decision, the Hamburg District Court ruled (September 27, 2024) that using copyrighted images in AI training falls under legal exceptions for scientific research, impacting how AI developers use protected content. The German court found that using copyrighted images in the AI training process was covered by exceptions in the legislation, notably for scientific research purposes and ruled against the plaintiff, who claimed that the defendant had infringed his copyright by using a photo in a dataset for training generative AI.

No international jurisdiction has yet altered patent protection terms specifically for AI-created inventions but scholars, policymakers, and legal experts continue to discuss whether current IP laws suffice or need updates to accommodate AI-generated works.

The intersection of AI and IP law is a dynamic frontier challenging traditional concepts. As AI systems become more sophisticated, capable of generating inventions with minimal human input, legal systems worldwide face the challenge of balancing innovation promotion with IP principles. Adapting to this new reality may require reexamining laws, crafting new legislation, and engaging in international cooperation to address the unique challenges AI poses in IP.