Ownership Of AI-Generated Cognitive Architecture And Behavioral Modeling Systems.
📌 1. Thaler (DABUS) — AI Inventorship in Patent Law (UK & Worldwide)
Cases: Thaler (Appellant) v Comptroller‑General of Patents, Designs and Trade Marks (UK)
Decision: Supreme Court of the United Kingdom (2023) — AI cannot be a legal inventor.
Facts
Dr. Stephen Thaler developed an AI system called DABUS that autonomously generated inventions (e.g., a fractal food container).
Thaler filed two patent applications listing DABUS as the inventor, arguing that as the owner of DABUS he should hold the patents.
Ruling
The UK Supreme Court held that:
Under the Patents Act 1977, an inventor must be a natural person. AI systems cannot be listed as inventors.
Ownership of an AI system (DABUS) does not automatically entitle the owner to the rights in the inventions created by that AI.
The applications were deemed withdrawn because no human inventor was named.
Legal Importance
Establishes that human inventorship is a statutory requirement — even if AI autonomously creates a novel invention.
Ownership cannot be inferred solely from owning the AI machine.
Global Echoes
The European Patent Office (EPO) rejected similar inventions claiming DABUS as inventor, holding that an AI cannot be inventor under the European Patent Convention.
The USPTO (United States) and many national offices similarly require a natural person inventor.
📌 2. US Copyright: Thaler’s AI‑Generated Artwork (DABUS Art)
Case: US Federal Appeals Court — 2025 ruling on AI art copyright.
Facts
Stephen Thaler sought copyright protection for artwork created entirely by his AI system, without human creative input.
Decision
The court upheld that copyright requires human authorship.
Works created solely by AI, without human creative contribution, cannot be copyrighted under U.S. law.
Implications
Even if an AI system produces a cognitive architecture design or behavioral model, ownership can only vest in a human author (or qualifying human contribution).
This aligns with U.S. Copyright Office guidance stating that AI output lacks human authorship if the AI executed core creative elements.
📌 3. Li v. Liu — Beijing Internet Court on AI‑Generated Works (China)
Case: Li v. Liu (2023) — Beijing Internet Court.
Facts
Li created an image using an AI model (Stable Diffusion) by inputting prompts and adjusting parameters at multiple stages.
A third party (Liu) published that AI‑generated image without permission.
Ruling
The court held that where a human:
Selects prompts;
Adjusts parameters;
Guides iterations;
Significance
This is a pro‑human ownership precedent in China: If human inputs meet originality criteria, the human (not AI) owns the copyright in AI‑generated output.
Signals that AI tools can be instruments — ownership follows human creative direction.
📌 4. Stability AI v. Getty Images — Copyright Training Data Dispute (UK)
Case: Getty Images v. Stability AI (UK High Court, 2025).
Background
Getty claimed Stability AI trained its image generating models on millions of copyrighted photos without permission.
The case involved complex claims of secondary copyright infringement, trade mark infringement, and database rights.
Outcome
The UK High Court rejected core copyright infringement claims because the AI model did not store or reproduce copyrighted works.
Stability AI was found to have committed trademark infringement where generated images contained Getty watermarks.
Ownership Insight
Even when AI generates output, courts will assess how underlying data was used and whether rights in training data or outputs are infringed.
Ownership disputes tend to focus on data usage rights and resulting reproductions, rather than treating AI as an originator.
đź“Ś 5. US Copyright Office Guidance + Related Case Trends
While not a single case, U.S. Copyright Office 2023–25 guidance and multiple court decisions reject copyright where AI alone produces creative output.
Key Factors
Works with significant human selection, arrangement, or creative control can qualify for copyright.
Pure machine output lacks traditional authorship and, hence, is not copyrightable.
This reinforces the human‑centric threshold in US copyright law.
📌 6. Additional International Insight — Divergent Jurisdictions
AI Ownership in Patent Law Globally
Some jurisdictions (e.g. Australia) initially allowed DABUS applications, but later courts reversed such allowances, agreeing only natural persons qualify as inventors.
Switzerland’s Federal Administrative Court similarly ruled that AI cannot be an inventor.
🧠Key Principles: How Courts Assign Ownership
âť— 1. Human Inventorship Required in Patents
In UK, US, EU, and many jurisdictions, only human inventors can be listed.
Even significant AI contributions do not overcome statutory definitions unless there’s meaningful human inventive input.
âť— 2. AI Output Copyright Depends on Human Creativity
Some jurisdictions (like China) will award copyright if human creative effort is evident in guiding AI.
In the US, pure AI output with negligible human input is not protected.
âť— 3. Data and Training Inputs Matter
Rights owners can challenge AI tools based on unauthorised use of copyrighted works during training, leading to secondary litigation on data rights.
âť— 4. Contractual Agreements Can Establish Ownership
In the absence of clear statutory rules, ownership and rights in AI‑generated systems (like cognitive architectures) often derive from contracts among developers, programmers, and users.
📌 Applying This to AI‑Generated Cognitive Architecture & Behavioral Modeling
For AI systems that generate complex models, systems, or architectures:
Patent Ownership
If an AI directly develops a novel architecture, courts will still require a human inventor to name — unless laws change.
Ownership of the AI machine alone does not confer ownership of the patent.
Copyright Ownership
Output (e.g., code, models, documentation) may be protected only if humans exercise creative judgment in setting design goals, architecture decisions, and model parameters.
Training Data Rights
Claims can be made if training data used by AI includes copyrighted or proprietary models/data without permission.
Contracts & Work‑For‑Hire
Organizations often assign ownership through contracts (e.g., employer owns outputs by employees). Courts uphold such agreements as long as parties agree on rights in advance.
đź“Ś Conclusion
Current legal doctrines worldwide emphasize human contribution, not AI autonomy, in determining ownership of AI‑generated innovations and works.
Patent systems require human inventorship — AI cannot be an inventor.
Copyright systems generally require human creative input to award ownership.
RELATED Blog
comments