The Artificial Inventor Project is an academic and legal research initiative that investigates the question of whether artificial intelligence systems can be legally recognized as inventors under existing patent law. Founded and led by Ryan Abbott, a professor of law and health sciences at the University of Surrey, the project seeks to clarify the legal status of AI-generated inventions through a coordinated series of test cases filed in patent offices and courts around the world. The initiative's central argument is that current patent laws, which typically require inventors to be natural persons, are outdated in the context of modern Artificial intelligence technologies and may stifle innovation by failing to protect inventions created autonomously by AI systems.
The project gained international attention in 2018 when it filed patent applications naming an AI system called DABUS (Device for the Autonomous Bootstrapping of Unified Sentience) as the sole inventor. DABUS was created by Stephen Thaler, a computer scientist and president of Imagination Engines Inc., who developed the system to generate novel ideas through a process of neural network-based associative memory. The applications covered a fractal-based beverage container and a device for attracting attention using a flashing light. By using DABUS as a test case, the project aimed to force patent offices to confront the legal question of whether an AI can be an inventor, regardless of whether the specific inventions were commercially valuable.
Legal Strategy and Test Cases
The project's legal strategy involved filing parallel patent applications in more than a dozen jurisdictions, including the United States, the United Kingdom, the European Patent Office (EPO), Australia, South Africa, Germany, Japan, and South Korea. The filings were deliberately identical in substance to ensure that any differences in outcomes could be attributed to variations in national patent laws and judicial interpretations. The project also engaged in extensive public outreach, publishing academic articles, giving presentations, and submitting amicus briefs in support of its position.
In South Africa, the project achieved a notable early success when the country's patent office granted the application in 2021, making it the first jurisdiction to formally recognize an AI as an inventor. However, this grant was largely procedural, as South Africa does not conduct substantive examination of patent applications. In Australia, the Federal Court initially ruled in favor of the project in 2021, but the decision was overturned on appeal in 2022. The United Kingdom's Supreme Court rejected the project's appeal in 2023, affirming that an inventor must be a natural person under UK law. The EPO also refused the applications, and subsequent appeals were dismissed.
Key Legal Arguments
The project's core legal argument rests on the interpretation of the term "inventor" in patent statutes. It contends that the law does not explicitly require an inventor to be human, and that the purpose of patent law - to encourage and reward innovation - is better served by recognizing AI as inventors when they autonomously generate patentable subject matter. The project also argues that failing to grant patents for AI-generated inventions creates a legal vacuum, leaving such inventions unprotected and potentially uncommercialized, which would harm economic growth and technological progress.
Opponents of the project, including several patent offices and courts, have countered that the concept of inventorship is inherently tied to human agency, moral rights, and the ability to understand and take responsibility for one's actions. They note that patent laws were drafted with human inventors in mind and that extending inventorship to AI would require legislative change rather than judicial reinterpretation. The project acknowledges this tension but maintains that courts have historically adapted patent law to new technologies, and that a narrow reading of "inventor" is not mandated by the statutory text.
DABUS and Its Inventor
Stephen Thaler developed DABUS as a system that uses multiple Neural network components to simulate creative processes. The system operates by creating a "chaotic" state of neural activity, then using a second network to evaluate and select outputs that meet certain criteria of novelty and utility. Thaler has described DABUS as a "creativity machine" that can generate ideas without human input, and he has used it to produce a range of concepts, from new types of food containers to emergency signaling devices. The project selected DABUS specifically because it could demonstrate autonomous generation, with no human contribution to the inventive step beyond providing the system with general goals.
Thaler's involvement has been both a strength and a point of contention for the project. While his system provided a concrete test case, some legal scholars have questioned whether DABUS truly operates autonomously or whether Thaler's programming and selection of outputs constitute a human inventive contribution. The project has responded that the standard for inventorship should focus on who or what conceived of the invention, and that DABUS's internal processes meet that threshold.
Academic and Policy Impact
Beyond the courtroom, the Artificial Inventor Project has stimulated a broader academic and policy debate about the role of AI in innovation. The project has published numerous papers in law reviews and intellectual property journals, and its findings have been cited in government consultations and reports from organizations such as the World Intellectual Property Organization (WIPO). In 2019, WIPO held a series of conversations on AI and intellectual property, and the project's arguments were prominently featured in those discussions.
The project has also influenced the development of soft law and policy recommendations. For example, the United States Patent and Trademark Office (USPTO) issued a request for comments on AI inventorship in 2019, and the project submitted a detailed response. While the USPTO ultimately concluded that current law does not permit AI to be named as an inventor, the agency acknowledged the importance of the issue and called for further study. The project continues to advocate for legislative reform, proposing model language that would allow AI to be listed as an inventor while retaining human ownership of patents.
Criticisms and Counterarguments
Critics of the project have raised several objections beyond the legal technicalities. Some argue that recognizing AI as inventors would devalue human creativity and undermine the moral foundations of the patent system. Others contend that the project's focus on autonomous AI is premature, as current AI systems, including Large language models and other Generative AI tools, are not truly independent but rely heavily on human-curated training data and prompts. The project responds that the law should be forward-looking and that the pace of AI development makes it prudent to address the question now rather than later.
Another line of criticism concerns the practical implications of AI inventorship. If an AI is named as an inventor, questions arise about who owns the patent, how to determine inventorship when multiple AI systems are involved, and how to handle liability for infringement. The project has proposed that ownership should default to the AI's owner or user, similar to how employers own patents for inventions made by employees. However, this proposal has not been universally accepted, and the project acknowledges that many details remain unresolved.
Global Divergence and Future Directions
The project's test cases have revealed a clear divergence in how different jurisdictions approach the issue. While South Africa's grant was a symbolic victory, most major patent offices have rejected the applications. The United States District Court for the Eastern District of Virginia ruled in 2021 that an AI cannot be an inventor under US law, and the Federal Circuit affirmed this decision in 2022. The Supreme Court declined to hear an appeal in 2023. In Germany, the Federal Patent Court allowed the application to proceed but required that the AI be listed as the inventor, with a human named as the applicant - a compromise that the project viewed as partially successful.
Looking ahead, the project plans to continue its advocacy through academic work, legislative proposals, and potential new test cases in jurisdictions that have not yet ruled on the matter. It also aims to engage with emerging technologies such as Deep learning and Transformer (architecture)-based systems, which are increasingly capable of generating patentable inventions in fields like drug discovery and materials science. The project's ultimate goal is to ensure that patent law evolves in a way that encourages innovation in the age of AI, rather than creating obstacles that could slow technological progress.
Broader Implications for Intellectual Property
The work of the Artificial Inventor Project has implications that extend beyond patents. It raises questions about copyright for AI-generated works, trade secrets, and the overall framework of intellectual property law. The project has argued that a consistent approach across all forms of IP is needed, and it has participated in discussions about AI-generated art and music. While the project's primary focus remains on patents, its findings have informed debates about whether AI should be granted legal personhood in other contexts, such as liability for autonomous vehicles or medical devices.
The project has also highlighted the need for international coordination. Because patent laws are territorial, an invention may be patentable in one country but not another, creating uncertainty for businesses operating globally. The project has called for harmonization through treaties and model laws, though it recognizes that achieving consensus among nations with different legal traditions is a long-term endeavor.
Conclusion
The Artificial Inventor Project has succeeded in placing the question of AI inventorship on the global legal agenda. Through its systematic test cases and scholarly output, it has forced courts and policymakers to articulate their reasoning and has exposed the limitations of existing laws. While the project has not achieved its primary goal of having an AI recognized as an inventor in a major jurisdiction, it has sparked a necessary conversation about how to adapt intellectual property systems to the realities of modern AI. As AI systems become more capable, the issues raised by the project are likely to become more pressing, and its work will remain relevant for years to come.