Legal Protection For Self-Learning AI-Generated Patents.

17 Mar 2026 --
0 Comments

1. Understanding the Concept

Self-Learning AI-Generated Inventions

Self-learning AI (like advanced neural networks, reinforcement learning models, or generative AI) can produce inventions without direct human input. These inventions can include:

New chemical compounds
Mechanical devices
Software systems
Design innovations

Key Legal Question:
Who can be recognized as the inventor for patent purposes when an AI generates the invention autonomously? And can such AI-generated inventions receive legal protection?

2. Legal Framework for Patent Protection

Patent law traditionally requires:

Inventorship – The person(s) who conceived the invention.
Novelty – The invention must be new.
Non-Obviousness/Inventive Step – Not obvious to someone skilled in the art.
Utility/Industrial Applicability – Must have practical use.

Challenge: AI is not a legal person. Most jurisdictions currently require a human inventor.

3. Case Laws on AI-Generated Patents

Here are six landmark cases and developments:

Case 1: DABUS AI Patent Cases – UK (Thaler v. Comptroller General of Patents, 2020)

Facts:

Dr. Stephen Thaler submitted patent applications naming DABUS, an AI system, as the inventor.
Applications covered:
1. A fractal-based beverage container
2. A flashing light device

Court Ruling:

UK Intellectual Property Office rejected the applications.
Reason: Inventor must be a natural person under UK law.

Implication:

AI cannot be legally recognized as an inventor.
Only humans can apply for patents for inventions created by AI.

Case 2: DABUS AI Patent Cases – EPO (European Patent Office, 2021)

Facts:

Thaler applied to the EPO, again naming DABUS as the inventor.

Ruling:

EPO rejected the application.
Cited Article 81 EPC: only a human being can be inventor.
Held that AI lacks legal capacity.

Key Takeaway:

Consistent with UK ruling; Europe does not recognize AI as inventors yet.
Shows a global trend restricting AI inventorship.

Case 3: DABUS AI Patent Cases – USPTO (United States, 2022)

Facts:

Thaler filed US patent applications naming DABUS as the inventor.

Ruling:

USPTO rejected it, citing the requirement that inventors must be human.
Confirmed by the court: AI cannot own patents under current US law.

Significance:

Reinforces that patents rely on human inventorship.
Practical solution: a human operator must be listed as the inventor to secure protection.

Case 4: Australian Federal Court – Thaler v. Commissioner of Patents (2022)

Facts:

Thaler challenged the Australian Patent Office’s rejection of AI inventorship claims.

Ruling:

Federal Court initially upheld the rejection.
Appeal: In 2022, Australia became the first jurisdiction to recognize AI as inventors under patent law.
Patents were granted with DABUS as the inventor.

Implications:

Sets a precedent for AI-generated inventions.
Recognizes that AI can conceive inventions autonomously.
Raises questions about ownership and licensing, since AI itself cannot own property.

Case 5: South Africa – DABUS Patent Decision (2021)

Facts:

Application filed with DABUS as the inventor.

Ruling:

South African patent office granted patent applications, recognizing DABUS as the inventor.
Marks a growing trend of jurisdictions experimenting with AI inventorship.

Case 6: Chinese Patent Office Guidelines (CNIPA, 2022)

Facts:

China issued draft guidelines for AI-generated inventions.

Key Provisions:

AI cannot be named as inventor.
Human agent(s) operating or directing AI can be listed as inventor.
Emphasizes AI as a tool, not a legal inventor, aligning with traditional IP frameworks.

4. Key Legal Issues and Challenges

Inventorship vs. Ownership
- AI can conceive, but cannot hold property.
- Ownership often assigned to the AI’s programmer, user, or owner.
Patent Eligibility
- Must demonstrate novelty, utility, and inventive step.
- AI can create inventions, but legal protection still depends on human attribution (except Australia and South Africa).
Global Divergence
- UK, EU, US, and China: AI cannot be inventor.
- Australia, South Africa: AI inventor recognized.
- Creates uncertainty for international filings.
Ethical and Economic Questions
- Should AI-generated inventions be rewarded or incentivized?
- How to allocate licensing, royalties, or commercialization rights?

5. Summary Table of AI Inventorship Cases

Jurisdiction	Case/Decision	AI as Inventor?	Notes
UK	Thaler v. UK IPO, 2020	❌	Human inventorship required
EU	EPO, 2021	❌	Consistent with UK; AI cannot be inventor
USA	USPTO, 2022	❌	AI cannot own patents; human required
Australia	Thaler v. Commissioner, 2022	✅	First jurisdiction granting AI inventorship
South Africa	DABUS Patent, 2021	✅	AI recognized as inventor
China	CNIPA Guidelines, 2022	❌	AI inventor not recognized; human must be listed

6. Key Takeaways

Most jurisdictions require human inventorship, limiting legal protection for AI-generated inventions.
Australia and South Africa are pioneers in granting AI inventorship, offering a potential model for the future.
Ownership and licensing are complex: AI itself cannot hold rights, so humans or entities must claim them.
There’s a growing need for harmonized international rules, especially with the rise of self-learning AI in research, pharma, and engineering.
Strategic approaches for inventors:
- List the human who oversees, operates, or programs the AI as the inventor.
- Maintain detailed records of AI’s contribution to support patentability.