How IP Law Will Evolve in the AGI Era: AI Inventions, AI Authorship, and the Regulatory Crossroads

As artificial general intelligence (AGI) comes into view, intellectual property law confronts foundational questions it has never had to answer. The concepts of “inventor” and “author” were designed on the assumption that humans are the primary subjects of creation and invention. In an era where AI systems independently generate inventions, text, images, and music, this assumption can no longer be taken for granted. This article analyzes the judicial decisions on AI inventorship, the copyright attribution problem for AI-generated works, the structural challenges AI poses to the patent system, the intersection with regulatory policy in the EU, the United States, and Japan, and three scenarios for how IP law may evolve over the next decade.

The AI Inventorship Problem: From DABUS to Today

The AI inventorship question is most closely associated with the litigation campaign of Dr. Stephen Thaler. Thaler filed patent applications in multiple jurisdictions naming his AI system DABUS (Device for the Autonomous Bootstrapping of Unified Sentience) as the inventor of two autonomous inventions.

Judicial Decisions by Jurisdiction

In the United States, the USPTO rejected the DABUS application, and the Federal Circuit affirmed in Thaler v. Vidal (43 F.4th 1207, August 2022), holding that “an inventor must be a natural person.” The Supreme Court denied certiorari in 2023, leaving the Federal Circuit’s ruling as the controlling authority. Under current U.S. law, an AI cannot be named as an inventor.

In the United Kingdom, the Supreme Court ruled in Thaler v. Comptroller-General of Patents, Designs and Trade Marks ([2023] UKSC 49, December 2023) that “an inventor must be a natural person or legal person, and an AI is neither,” and declined to register the application. The decision expressly left open the separate question of whether an AI’s owner or developer could obtain a patent as inventor in cases where an AI performs the inventive act.

The EPO also rejected the DABUS application, with the Technical Board of Appeal maintaining in 2020 (T0844/18) that an inventor must be a natural person. South Africa granted the DABUS patent in 2021, and Australia’s Federal Court initially recognized AI inventorship (Thaler v. Commissioner of Patents [2021] FCA 879), but the Full Federal Court subsequently reversed that decision.

What the Divergence Means

The jurisdictional divergence is not merely a technical legal dispute. As AI increasingly leads the inventive act in practice, the “natural person inventor” requirement is creating a growing gap between doctrine and reality. As of 2026, no major patent office recognizes AI as an inventor, but patent applications by human inventors who used AI as a tool remain permissible—and establishing practical standards for evaluating such “human + AI” collaborative inventions has become an urgent policy priority.

Copyright Attribution for AI-Generated Works

The U.S. Copyright Office’s Position

In March 2023, the U.S. Copyright Office determined that images autonomously generated by an AI cannot be registered (Re: Second Request for Reconsideration for Refusal to Register SURYAST, SR# 1-7150-14432). The Office has consistently maintained that “copyright vests in human creativity.”

In February 2023, in response to a registration application for Zarya of the Dawn—a graphic novel created using the Midjourney image-generation AI—the Office declined copyright protection for the AI-generated images while granting protection for the human-authored text and the selection and arrangement of visual elements. This ruling provides a framework for assessing protection scope in “human + AI” collaborative works.

Japan and the EU

Japanese copyright law defines a work as “a creative expression of thoughts or emotions” (Article 2, Paragraph 1) and an author as “a person who creates a work” (Article 2, Paragraph 2). Because AI-generated content is not produced by a subject possessing “thoughts or emotions,” the prevailing view is that such content does not qualify for copyright protection. The Agency for Cultural Affairs has indicated a flexible interpretation, however: the extent of copyright protection may vary depending on the degree of human involvement in producing the output, with the level of “creative contribution” functioning as the key criterion.

The EU has not yet adopted uniform rules on copyright attribution for AI works. The EU AI Act (2024) does not directly regulate copyright in AI-generated content, and while the European Parliament has expressed reluctance to grant copyright to content generated by autonomous AI, formal EU-level consensus has not yet been codified in legislation.

The Structural Challenges AI Poses to Patent Law

The Non-Obviousness Standard in Crisis

Non-obviousness (35 U.S.C. §103 in the U.S.; EPC Article 56 in Europe; Article 29, Paragraph 2 of Japan’s Patent Act) requires that an invention “would not have been obvious” to a person having ordinary skill in the art (PHOSITA) at the time the invention was made. As AI advances, technical problems that once took skilled engineers years to resolve are being solved in hours. The rapid elevation of the PHOSITA’s capability level threatens to hollow out the non-obviousness standard as conventionally applied.

Since 2024, the Federal Circuit has begun to engage with the question of how to assess PHOSITA capability in an era where AI-assistance tools are widely available, and USPTO examiners are being called upon to grapple with the same issue in practice.

The Explosion of Prior Art

AI-related paper output—including arXiv preprints—has grown at explosive rates. AI/ML submissions to arXiv exceeded 500,000 in 2024 alone. The cost of identifying genuinely novel and non-obvious inventions within this enormous body of prior art is rising sharply for patent examiners. The USPTO has been introducing AI-assisted examination tools, but examination capacity has not yet kept pace with the growth in the prior art base.

The Regulatory Intersection: EU AI Act, U.S. Executive Orders, and Japan’s AI Strategy

EU AI Act (2024)

The EU AI Act (Regulation (EU) 2024/1689) entered into force in August 2024. It classifies AI systems into four risk tiers and mandates transparency, accountability, and human oversight for high-risk systems. On the IP intersection, the Act requires providers of general-purpose AI models (equivalent to GPT-class systems) to publish “sufficiently detailed summaries” of content used for training (Article 53). This requirement directly affects AI companies’ training data governance and copyright clearance practices.

United States: Executive Orders and Congressional Activity

The Biden Administration’s Executive Order on AI Safety (Executive Order 14110, October 2023) directed a whole-of-government approach to AI safety, transparency, and IP protection. IP-related directives included instructing the USPTO and the Copyright Office to review their policies on AI inventorship and AI authorship. In response, the USPTO revised its AI inventorship guidance in 2024.

At the congressional level, multiple AI copyright proposals have been introduced, but as of April 2026, no comprehensive AI copyright statute has been enacted. Bipartisan consensus has proven difficult to build, with AI companies, content rights holders, and the academic community supporting divergent legislative directions.

Japan’s AI Strategy and IP Policy

Japan established an AI Strategy Council in 2023 to advance AI utilization and IP system development in parallel. The Cabinet Office’s Intellectual Property Strategy Headquarters published an AI-era IP policy package in 2024, articulating three priorities: (1) clarifying rights clearance for training data (refining Article 30-4 application); (2) establishing copyright attribution rules for AI-generated works; and (3) pursuing international coordination on the AI inventorship problem. Japan’s comparatively permissive legal framework for AI training functions as a competitive advantage for its AI industry.

Three Scenarios for IP Law in 2036

The IP system in the AGI era may diverge along several trajectories. Three scenarios looking toward 2036 are outlined below.

Scenario A: Adaptive Evolution — Incremental Institutional Reform

In this scenario, patent offices, copyright offices, and courts gradually calibrate existing standards to accommodate AI without large-scale legislative reform. For patents, practical guidelines for “AI-assisted inventions” take shape, with protection available where meaningful human creative contribution can be demonstrated. For copyright, protection standards for “human + AI” works are established based on the degree of human selection and judgment involved. This scenario represents the natural extension of current trends and is the most probable outcome.

Scenario B: Institutional Rupture — Radical Reform for the AGI Era

In this scenario, the realization of AGI—systems capable of autonomous scientific discovery, technical invention, and artistic creation—triggers recognition that existing institutions have fundamentally broken down, prompting radical legislative overhaul. New patent provisions recognize “AI inventors,” establishing a framework under which AI developers own the resulting patents. Copyright law creates a special category of “AI-generated works” eligible for limited-term protection (for example, 10 years). This scenario requires strong political will and international coordination and is currently of low probability, though technological breakthroughs could accelerate its arrival.

Scenario C: Return to the Commons — Contraction of Protection

In this scenario, the rapid proliferation of AI reduces the practical effectiveness of IP protection to the point where open-source and Creative Commons-style sharing frameworks become dominant. The cost-benefit analysis of patent filing turns negative as technology commoditizes faster than patents can be enforced, shifting the primary protection mechanism for AI inventions from patents to trade secrets. As the cost of generating copyrighted content approaches zero, debates over shorter protection terms and reduced exclusive rights intensify. This scenario fundamentally transforms the IP investment return model for AI companies and forces a broader business model transformation.

Conclusion: A Non-Linear Future for IP Institutions

The future of IP law in the AI and AGI era does not converge on a single linear scenario. The pace of technological progress, regulatory decisions by key jurisdictions, the outcome of major litigation, and shifting social and ethical values will interact in complex ways to determine the direction of institutional change.

One thing can be said with certainty. IP institutions have historically adapted to technological change with a lag—but the pace of AI innovation is faster than any prior technology, and its generality broader than any prior invention. Institutional designers face an unprecedented demand for speed and flexibility in adaptation. In the AI IP war’s final reckoning, the ultimate winner may not be the company with the best technology, but the society that constructs the wisest institutions.

---

This is Part 5 (the final installment) of the “AI Patent War of 2026” series. See also Parts 1 through 4.