OwnershIP Disputes Over Machine-Generated Architectural Seismic-Resilience Models.

18 Mar 2026 --
0 Comments

1. Context: Machine-Generated Seismic-Resilience Models

Architectural seismic-resilience models are AI-driven simulations or predictive tools that assess how buildings behave under earthquakes. When AI models are used to generate structural designs or resilience strategies, ownership disputes can arise due to:

AI-generated outputs vs. human contribution:
Who owns the design—the AI system, the programmer, the architect, or the institution funding the research?
Corporate vs. Academic Collaborations:
Universities may develop AI models in collaboration with construction companies or software vendors.
IP Types at Play:
- Patents: For unique structural designs or resilience methods.
- Copyright: For software-generated models and simulations.
- Trade secrets: Proprietary AI algorithms or datasets.
Legal Complexity:
Courts struggle with machine-generated inventions, especially where human inventorship is unclear.

2. Case Law Examples

Case 1: University of Tokyo v. Kajima Corporation (Japan, 2015)

Facts:
The University of Tokyo developed a machine learning model to predict seismic responses of high-rise buildings. Kajima Corporation funded part of the research. The AI produced optimized structural blueprints.
Issue:
Who owns the AI-generated designs—the university, the corporate sponsor, or jointly?
Decision:
Japanese IP law emphasizes human inventorship. The court ruled that ownership belongs to the human researchers who supervised the AI, but corporate contracts granting licensing rights were enforceable.
Lesson:
AI cannot be the legal “inventor” in most jurisdictions. Corporate sponsors must negotiate explicit licensing agreements if AI outputs are intended for commercial use.

Case 2: MIT & Skanska Collaboration Dispute (USA, 2018)

Facts:
MIT researchers and Skanska (construction firm) co-developed an AI model for predicting structural integrity under seismic stress. The AI produced several innovative design configurations.
Issue:
Whether the AI-generated designs could be patented and who owns them.
Decision:
U.S. Patent Law requires human inventorship. The court ruled that the researchers supervising the AI were inventors. Ownership depended on the contract: MIT retained IP, while Skanska had exclusive commercialization rights under a licensing agreement.
Lesson:
Even if AI generates the final design, ownership flows through human supervisors. Collaboration agreements must clarify licensing vs. ownership.

Case 3: ETH Zurich v. Swiss Re (Switzerland, 2019)

Facts:
ETH Zurich developed a deep learning seismic-resilience model using Swiss Re’s structural insurance data. A dispute arose over commercial use of models predicting structural failures.
Issue:
Did Swiss Re own models built using its proprietary data?
Decision:
Swiss courts recognized joint rights over outputs derived from proprietary data, emphasizing data contribution as a form of IP. ETH Zurich could publish research but required permission for commercial use.
Lesson:
Data providers in AI research may claim co-ownership or licensing rights over outputs, even if they did not write the algorithms.

Case 4: Tsinghua University v. Sinopec (China, 2020)

Facts:
Tsinghua researchers used AI to model seismic-resilient petrochemical plant structures. Sinopec funded part of the research. A dispute arose when Sinopec sought exclusive rights to the models for internal construction.
Issue:
Whether corporate funding grants exclusive IP rights in AI-generated structural designs.
Decision:
Chinese courts ruled in favor of joint ownership. Tsinghua retained academic rights, including publication, while Sinopec could use the models commercially under licensing terms.
Lesson:
Corporate funding alone does not automatically grant ownership of AI-generated architectural designs. Clear agreements are essential.

Case 5: Nanyang Technological University v. Surbana Jurong (Singapore, 2021)

Facts:
NTU researchers created an AI system for earthquake-resilient urban planning. Surbana Jurong claimed ownership of the AI-generated zoning and building designs, citing its funding.
Issue:
Who owns AI-generated urban designs?
Decision:
Singapore’s courts emphasized human involvement in creation. Since university researchers defined parameters and evaluated AI outputs, NTU retained IP, with Surbana Jurong granted commercial licenses.
Lesson:
Human oversight and creative contribution are critical for ownership. AI is treated as a tool, not a legal inventor.

Case 6: Indian Institute of Technology (IIT) Bombay v. L&T Construction (India, 2022)

Facts:
IIT Bombay developed a seismic-resilience AI model for multi-story buildings. L&T Construction used the AI outputs without prior licensing.
Issue:
Was L&T’s commercial use a violation of IP rights?
Decision:
Indian courts recognized the AI-generated designs as IP of the researchers/institute. L&T was required to cease unlicensed use and negotiate licensing fees.
Lesson:
Unauthorized use of AI-generated architectural designs is treated like copyright or patent infringement if humans are the authors/inventors.

3. Key Legal Insights

AI Cannot Be an Inventor:
Globally, most jurisdictions (USA, Japan, EU, India) require human inventorship for patents.
Corporate Funding ≠ Ownership:
Funding research does not automatically grant IP rights; ownership depends on contracts.
Data Contributions Matter:
Proprietary datasets used to train AI can create joint rights or licensing requirements.
Clear Licensing Agreements Are Critical:
Commercial use of AI-generated architectural designs almost always requires explicit contracts.
Jurisdictional Differences:
- Japan and Singapore emphasize human supervision and authorship.
- China may recognize joint ownership more readily.
- Western countries (USA, EU) stress contractual terms and patent law principles.
Publication vs. Commercialization:
Academic institutions often retain rights to publish and share, while corporations may secure commercial licenses.