Can AI-Generated Inventions Be Patented? Global Legal Developments

The DABUS Saga: A Global Litmus Test for AI Inventorship

The most prominent and widely discussed case concerning AI inventorship is undoubtedly the DABUS (Device for the Autonomous Bootstrapping of Unified Sentience) saga. Developed by Dr. Stephen Thaler, DABUS is an AI system that Thaler claims independently conceived two inventions: a food container based on fractal geometry and a flashing light for attracting attention. Thaler sought to list DABUS as the inventor on patent applications in numerous jurisdictions, leading to a series of landmark decisions.

Jurisdictional Responses to DABUS

The outcomes of the DABUS applications have been strikingly varied, highlighting the disparate approaches taken by patent offices and courts worldwide:

• South Africa (2021): Made history by being the first country to grant a patent listing DABUS as the inventor. The South African Companies and Intellectual Property Commission (CIPC) approved the application without substantive examination specifically on the inventorship question, focusing instead on formal requirements.
• Australia (2021-2022): Initially, a Federal Court judge ruled in favor of DABUS as an inventor, stating that "an inventor can be an AI system or device." However, this decision was overturned by the Full Federal Court in 2022, which unanimously concluded that an inventor must be a natural person under Australian patent law. The High Court of Australia later denied leave to appeal, effectively confirming the "human inventor" requirement.
• United Kingdom (2020-2023): The UK Intellectual Property Office (IPO) rejected the applications, a decision upheld by the High Court and the Court of Appeal. The Supreme Court of the United Kingdom, in December 2023, unanimously dismissed Dr. Thaler's appeal, affirming that only a natural person can be an inventor under the UK Patents Act 1977. The court emphasized that a "person" in the context of inventorship refers to a human being.
• United States (2020-2023): The United States Patent and Trademark Office (USPTO) rejected the applications, a decision affirmed by the Eastern District of Virginia and subsequently by the Court of Appeals for the Federal Circuit (CAFC) in 2022. The CAFC held that the Patent Act "requires that an inventor be an individual," meaning a natural person. The Supreme Court denied certiorari in 2023, leaving the CAFC's decision as the standing precedent.
• European Patent Office (EPO) (2020): The EPO Boards of Appeal upheld the EPO's earlier rejection, stating that an inventor must be a natural person with legal personality. The EPO emphasized that the European Patent Convention (EPC) requires the inventor to be a "person" and that naming a machine would not meet this requirement.
• Germany (2023): The German Federal Patent Court allowed Thaler to list himself as the inventor, but with an accompanying statement that DABUS contributed to the invention. This was a pragmatic workaround rather than a recognition of AI inventorship.

"The DABUS cases unequivocally demonstrate a global consensus, with the notable exception of South Africa, that current patent laws are not equipped to recognize AI systems as legal inventors. This highlights a fundamental tension between technological advancement and established legal principles."

The Legal and Philosophical Underpinnings

The core of the debate revolves around several key legal concepts:

• Inventorship: Traditionally, inventorship is tied to the concept of "conception," the formation in the mind of the inventor of a definite and permanent idea of the complete and operative invention. Can an AI "conceive" in this human sense?
• Legal Personality: Patent rights are granted to legal entities. AI systems currently lack legal personality, meaning they cannot own property, enter contracts, or be sued.
• Policy Rationale: Patent systems incentivize human ingenuity by granting exclusive rights. If an AI is the inventor, who receives the incentive? The AI's owner, its programmer, or no one?

Global Policy Responses and Emerging Frameworks

While courts have largely adhered to existing legal definitions, patent offices and governments are actively exploring policy adaptations.

United States: Call for Stakeholder Input

The USPTO has been proactive in soliciting public comments on AI and inventorship. In 2023, it issued a request for comments on "Artificial Intelligence and Inventorship," seeking input on:

• Whether current patent laws are sufficient or require modification.
• How to determine inventorship when AI significantly contributes.
• The impact on patent quality and enforcement.

This indicates a recognition that while current law rejects AI as an inventor, the underlying policy considerations are far from settled.

European Union: Expert Group Reports and Consultations

The European Commission’s Expert Group on AI has published several reports touching upon intellectual property rights, including inventorship. While no legislative changes are imminent, discussions revolve around:

• The "Human-Centric" Approach: Most discussions favor maintaining a human inventor but acknowledging AI's role through ownership or co-inventorship with human collaborators.
• Ownership of AI-Generated Works: Proposing that the owner or operator of the AI system could be considered the owner of the resulting IP.
• New Forms of Protection: Exploring whether existing IP frameworks are adequate or if new categories of protection are needed for purely AI-generated outputs.

China: Growing Importance of AI in Patent Filings

China is a global leader in AI patent filings. While its patent law, like most, requires a natural person as an inventor, the sheer volume of AI-related inventions means that the role of AI in the inventive process is a practical concern.

• Practical Approach: Chinese patent examiners are likely to focus on identifying human contribution in AI-assisted inventions, rather than directly challenging the human inventorship claim unless explicitly stated otherwise.
• Government Initiatives: The Chinese government has released White Papers and strategic plans on AI development, which often include discussions on IP protection, indicating a readiness to adapt policies as AI capabilities advance.

World Intellectual Property Organization (WIPO): Global Dialogue

WIPO has initiated a series of "Conversations on IP and AI," bringing together experts from various fields to discuss the implications of AI for IP policy. These discussions cover:

• Inventorship: The definition of an inventor and whether AI can be recognized.
• Authorship: Similar questions for copyright.
• Ownership: Who owns the IP generated by AI.
• Data Issues: The use of copyrighted/patented data for training AI.

"The global legal community is not ignoring AI's impact; rather, it is in a reactive phase, interpreting existing laws, while simultaneously engaging in proactive policy discussions to shape future frameworks. The focus is shifting from 'can AI be an inventor?' to 'how do we properly attribute and incentivize innovation in an AI-assisted world?'"

Practical Advice for AI-Assisted Inventions

Given the current legal landscape, where AI is largely excluded from inventorship, here are practical strategies for innovators leveraging AI:

1. Document Human Contribution Meticulously

• Identify the Human Spark: Even if an AI generates numerous potential solutions, the human who frames the problem, selects the AI's inputs, interprets its outputs, refines the solution, or recognizes its inventiveness, contributes to conception. Document these human decisions and interventions.
• Problem Formulation: The human defining the problem that the AI is tasked to solve is a crucial inventive step.
• Algorithm Design/Selection: The human choosing or designing the specific AI algorithms or models used in the inventive process.
• Data Curation and Training: While AI learns from data, the human selection, preparation, and labeling of training data can be an inventive act.
• Result Interpretation and Selection: The human activity of reviewing AI-generated outputs, identifying the novel and non-obvious elements, and selecting which outputs to pursue.
• Refinement and Implementation: The human act of taking an AI-generated concept and further refining it, or developing it into a practical, implementable invention.

2. Focus on "Human-in-the-Loop" Inventorship

• Co-Inventorship: If an AI significantly aids in the invention, consider listing the human operators, programmers, or researchers who guided the AI or interpreted its results as co-inventors.
• No Pure AI Inventorship Claims: Avoid claiming an AI system as the sole inventor in jurisdictions that do not recognize it (which is most of them). This will almost certainly lead to rejection and delays.

3. Review Patent Office Guidelines

• Stay Updated: Patent offices, particularly the USPTO, EPO, and national offices, regularly issue guidance or solicit comments on AI-related inventions. Consult these official documents for the latest interpretations.
• Example: USPTO Inventorship Guidance (forthcoming): Expect future guidance from the USPTO on how to assess inventorship in AI-assisted inventions, likely emphasizing the "significant contribution" test for human co-inventors.

4. Consider Ownership and Assignment

• Clear Agreements: Ensure clear contractual agreements regarding ownership of inventions created with AI tools, especially between developers, users, and employers.
• Employer-Employee Relationships: Standard employment agreements typically assign inventions to the employer. This principle generally extends to AI-assisted inventions made within the scope of employment.

5. Strategy for Novelty and Non-Obviousness

• Prior Art Search: AI can be a powerful tool for conducting prior art searches, but the analysis and interpretation of results still largely fall to human experts.
• Demonstrate Inventive Step: Be prepared to articulate how the human contribution, even if AI-assisted, satisfies the requirements of novelty and non-obviousness over existing prior art. The AI itself cannot be used as a basis for identifying the "person having ordinary skill in the art" (PHOSITA).

6. Explore Alternative Protections

• Copyright: For purely creative works generated by AI (e.g., music, art), copyright laws present similar inventorship/authorship challenges. However, for certain AI outputs, copyright protection might be viable if a human has made sufficient creative choices.
• Trade Secrets: For underlying AI algorithms, training data, or processes that are not publicly disclosed, trade secret protection can be a strong alternative, provided confidentiality is maintained.

"The key to patenting AI-assisted inventions lies in demonstrating a clear and significant human intellectual contribution to the conception of the invention. While AI can amplify human capabilities, the current legal framework insists on a human 'mind' behind the inventive spark."

Conclusion

The debate surrounding AI inventorship is far from over. While the DABUS decisions have largely cemented the "human inventor" requirement under current laws, they have also ignited a global dialogue about the future of intellectual property in an AI-driven world. Innovators must proceed pragmatically, focusing on meticulously documenting human contributions and adhering to existing legal frameworks, while simultaneously advocating for and adapting to evolving policy landscapes. The journey toward a comprehensive and equitable framework for AI-generated inventions is just beginning.

Frequently Asked Questions

### Q1: Can an AI system be listed as an inventor on a patent application in any country?

Currently, only South Africa has officially granted a patent listing an AI system (DABUS) as an inventor. Most other major jurisdictions, including the US, UK, EU, Australia, and China, have explicitly rejected this possibility, maintaining that an inventor must be a natural person.

### Q2: If an AI significantly helps create an invention, who should be listed as the inventor?

In most jurisdictions, the human(s) who made a significant intellectual contribution to the conception of the invention should be listed as the inventor(s). This includes individuals who formulated the problem, designed the AI's parameters, interpreted the AI's outputs, or refined the AI-generated solution. It's crucial to document these human contributions meticulously.

### Q3: What is the primary reason patent offices reject AI as an inventor?

The primary reason is that existing patent laws define an "inventor" as a natural person (a human being) capable of "conception" – forming a definite and permanent idea of the invention in their mind. AI systems currently lack legal personality and the cognitive capacity to "conceive" in the human sense required by these laws.

### Q4: Will patent laws change to recognize AI as an inventor in the future?

It's uncertain. While current laws largely reject AI inventorship, there is a growing global discussion among policymakers, patent offices, and legal experts about whether and how patent systems should adapt. Potential changes could include new categories of inventorship, modified ownership rules, or even entirely new intellectual property rights for AI-generated output. However, any such changes would likely be complex and take significant time to implement due to the fundamental policy shifts required.