Patentability of Artificial Intelligence-Generated Inventions: Emerging Challenges in Patent Law

Artificial intelligence (AI) has transformed nearly every major industry, from healthcare and biotechnology to software development and autonomous systems. With AI now capable of generating solutions, designs, and even novel products with minimal human intervention, patent regimes worldwide face a critical question: Can AI be an inventor? This issue, once theoretical has become one of the most debated topics in modern patent law. As patent systems are traditionally built around human creativity, AI-generated inventions challenge the foundational principles of inventorship, inventiveness, and ownership. This article examines the evolving legal landscape of AI-generated inventions, the challenges they pose, and the direction patent law may take in the coming years.

Understanding AI-Generated Inventions

AI-generated inventions can be broadly categories into two groups:

1. AI-assisted inventions – where AI helps a human inventor but does not independently conceive the invention. For example, AI software that suggests modifications to a drug molecule while the scientist makes the final determination.

2. AI-autonomous inventions – where AI independently generates the concept, design, or solution without significant human input. Famous examples include inventions generated by the AI system DABUS, such as a fractal container design and an emergency beacon.

While patent systems worldwide can fairly accommodate AI-assisted inventions, AI-autonomous inventions are where legal uncertainty emerges, as most laws require inventors to be natural persons.

Inventorship Requirements in Patent Law

A core requirement under nearly all patent systems is that an invention must be attributed to a human inventor. This principle is embedded in:

• The US Patent Act, which defines inventors as “individuals.”

• The European Patent Convention (EPC), which requires inventors to be “natural persons.”

• The Indian Patents Act, 1970, which uses the term “person,” historically interpreted as a human being.

The rationale for requiring a human inventor is rooted in the philosophy that patents reward human ingenuity. Patent systems were created to encourage human creators to innovate by offering time-limited monopolies.

However, when AI autonomously creates an invention, the question arises: Who is the inventor when no human can truthfully claim to have conceived the idea?

Landmark Cases: The DABUS Decisions

The debate intensified with the filing of patent applications naming the AI system DABUS as the inventor. Courts and patent offices across the world ruled on the matter:

United States

The US Court of Appeals held that only humans can be inventors. The USPTO rejected the application on grounds that “individual” means a natural person.

United Kingdom

The UK Supreme Court affirmed the same view that the Patents Act requires a human inventor.

European Union

The EPO also rejected the application, citing that inventorship must be attributed to a natural person.

South Africa

South Africa was the first country to grant a patent with DABUS named as inventor but this was due to a non-examination system, not a substantive legal decision.

Australia

Initially, a federal judge ruled AI could be an inventor, but the decision was overturned on appeal.

These global decisions show a consistent trend: patent law currently does not accommodate AI as an inventor.

Challenges Posed by AI-Generated Inventions

(a) Inventorship and Ownership Issues

If AI cannot be an inventor, the question becomes: Who owns the output of AI-generated inventions?

Possible candidates include:

• the human programmer,

• the user operating the AI,

• the company owning the AI system.

But ownership typically flows from inventorship in patent law, and if no human inventor exists, the entire system struggles.

(b) Inventive Step and Non-Obviousness

AI can generate solutions after processing large datasets using algorithms. Critics argue:

• If AI does most of the work, should the output be considered non-obvious?

• Should the patent system reward human inventors for inventions they did not “conceive”?

The inventive step test may need re-examination to account for AI’s capability to produce highly optimized solutions.

(c) Disclosure Requirements

Patent law requires the applicant to fully disclose how the invention works. But AI models like deep neural networks often operate as "black boxes". This raises concerns:

• Can applicants adequately disclose the invention?

• Should AI algorithms themselves be disclosed?

(d) Proliferation of Patents and Evergreening

AI can generate millions of variations of compounds or designs. Patent offices may face:

• floods of applications,

• low-value patents,

• increased risk of patent thickets and harmful monopolies.

Policy Responses Around the World

Governments and patent offices are actively studying the issue.

United States

The USPTO released guidance acknowledging AI's growing role but reaffirmed human inventorship. It is considering potential reforms for AI-assisted inventions but remains cautious about granting rights to fully autonomous AI systems.

European Union

EPO guidelines emphasize the human-centered approach but recognize AI as an important tool in several industries. The EU AI Act also influences discussions.

India

India has not yet taken a formal stance but follows the global trend of human inventorship. With India's growing AI economy, discussions on AI-generated patents are likely to intensify.

WIPO (World Intellectual Property Organization)

WIPO is conducting consultations to develop international frameworks, recognizing the need for harmonization across jurisdictions.

The Way Forward: Possible Legal Approaches

As AI continues to evolve, patent law may need new frameworks. Possible approaches include:

(a) Human-Centric Approach (Status Quo)

Only humans are inventors; AI is treated as a tool. This protects existing legal structures but may undervalue AI’s autonomy.

(b) AI as Joint Inventor

Some scholars argue for recognizing AI as a co-inventor alongside humans.

(c) AI as an Inventor (Future Scenario)

This would require major legal reforms and raises philosophical questions about personhood.

(d) Creating a Separate Regime for AI-Generated Innovations

Instead of patents, AI creations could be rewarded through:

• sui generis intellectual property rights,

• utility certificates,

• shorter-term protection systems.

Conclusion

AI-generated inventions challenge the very foundations of patent law particularly inventorship, inventive step, and ownership. While patent regimes worldwide currently limit inventorship to human beings, technological advancements may force future reconsideration. If AI continues to evolve into a truly autonomous creator, legal systems must adapt to balance innovation incentives, ethical considerations, and economic realities.

A harmonized international framework will be essential to address these challenges effectively. Until then, patent law will continue to grapple with the question: How do we protect inventions that are not created by humans?