OwnershIP Disputes For Global AI Trade-Flow Prediction Systems.
🧠 1 | Thaler v. United States (AI as Inventor / Copyright Owner)
Jurisdiction: United States (Federal Courts)
Legal Issue: Can an AI system be recognized as the inventor or author for purposes of patent or copyright ownership?
Ruling / Outcome:
- Lower courts and the U.S. Court of Appeals in the D.C. Circuit consistently held that AI cannot be an inventor or author under U.S. law because legal personhood and authorship/inventorship require a human being.
- The U.S. Supreme Court refused to hear an appeal on whether AI‑generated art can be copyrighted, effectively affirming the lower court’s stance that only humans can own copyright.
Why It Matters for Predictive AI:
AI models used to forecast trade flows — even if they generate novel insights or analytical products — cannot themselves own IP. The legal effect is that ownership rights must be traced back to human designers, users, or contractual arrangements (e.g., employers, licensors).
This case sets the global precedent that mere autonomous output by software does not vest IP rights in the machine, and without human authorship or contractual assignment, the rights default to humans or corporate entities.
🧾 2 | Concord Music Group Inc. v. Anthropic PBC (AI Copyright / Fair Use Litigation)
Jurisdiction: United States (Northern District of California)
Legal Issue: Whether AI training on copyrighted materials without express permission constitutes infringement or is protected by fair use.
Details & Outcome:
- Major music publishers filed suit alleging Anthropic’s AI used copyrighted song lyrics to train its model, which then generated similar outputs.
- Plaintiffs argue AI training is not a fair use and competes directly with original works.
- The court is being asked to rule that Anthropic’s models cannot invoke fair use.
Legal Insights:
- This has major implications for trade‑flow prediction AI: such systems are often trained on proprietary financial, trade, or logistics data.
- If the input data is owned by others (e.g., customs datasets, shipping manifests, subscription financial feeds), similar disputes could arise over improper copying or reuse without licensing.
📘 3 | Bartz et al. v. Anthropic PBC (AI Training & Copyright Fair Use)
Jurisdiction: United States (Federal Court)
Legal Issue: Whether training AI on copyrighted books is fair use.
Outcome:
- Judge William Alsup ruled training on lawfully obtained books may be fair use under U.S. law, as such use is “transformative.”
- But allegations about use of pirated copies remain part of the trial record.
Relevance:
- Predictive systems trained on proprietary databases (e.g., trade flow histories, shipment records) could provoke similar disputes if the legality of how data was sourced or used is contested.
🎨 4 | Li v. Liu (Beijing Internet Court — AI Content Ownership)
Jurisdiction: China
Legal Issue: Who owns the copyright in AI‑generated images?
Outcome:
- The Beijing Internet Court ruled that an AI‑generated image can be protected by copyright, and that the human user who provided the prompts and made creative choices is the copyright owner, not the AI tool.
Takeaways for Predictive AI:
- This is a non‑U.S. example of assigning ownership to humans, but only when there is meaningful intellectual input.
- Courts abroad may adopt similar principles for concrete AI outputs tied to human direction, even in analytical contexts — such as trade forecasting models or reports — if human architects have provided direction and selection.
📸 5 | Andersen, McKernan & Ortiz v. Stability AI, Midjourney & DeviantArt
Jurisdiction: U.S. Federal Court
Legal Issue: Whether training AI on scraped images without permission is copyright infringement.
Outcome:
- Artists claimed the training of over billions of images violated their rights; the judge allowed parts of the complaint to survive, indicating that training data sourced without consent can be actionable.
Relevance:
- Predictive AI—especially those modeling global trade—often uses large scraped or purchased datasets. This case shows that lack of licensing or permission for training data can become the core of ownership/infringement litigation.
💡 6 | Anthropic Settlement (Authors vs AI Firm — Pirated Books)
Jurisdiction: United States (Class Action Resolution)
Legal Concept: Settlement for alleged copyright infringement of training data.
Outcome:
- Authors settled with Anthropic for $1.5 billion over use of copyrighted books obtained through piracy for AI training.
Legal Importance:
- This shows the economic stakes when disputes over unauthorized data use escalate. Predictive AI vendors that rely on potentially copyrighted proprietary sources without proper licensing could face similar large damages.
🧠 Key Legal Themes (Applicable to Predictive AI & Trade‑Flow Systems)
👥 Human Authorship / Inventorship
Most jurisdictions require human involvement for IP protection — including copyrights or patents — meaning that AI cannot own output, and humans or entities must claim inventive/creative roles.
📊 Training Data Ownership
Disputes often focus on whether datasets — especially proprietary ones — were used lawfully. Unauthorized data scraping or purchase without licensing is increasingly litigated.
📝 Contractual vs. Statutory Rights
Many disputes turn on licensing agreements between data owners and AI developers — not just statutory copyright rules. Predictive systems with proprietary data must have clear licensing to avoid ownership disputes.
⚖️ Fair Use / Transformative Use
In U.S. copyright law, courts are carving out fair use defenses for certain AI training, but this is fact‑specific. The nature of the use (transformative or commercial) matters greatly.
📌 How These Apply Specifically to Global Trade‑Flow Prediction AI
While most case law to date involves creative or general generative AI, the legal principles translate directly:
- A predictive model trained on third‑party proprietary trade data could face IP claims if training data was copied without authorization.
- If model outputs closely resemble proprietary reports or forecasts, owners may allege infringement or derivative rights violation.
- Ownership disputes often hinge on contracts, licensing terms, and whether human strategy/innovation contributed to the model’s design and use.
- Clear documentation of data rights and human contributions is becoming essential to defend AI output ownership claims.
📌 Practical Lessons for Developers & Users
| Issue | Legal Risk | Legal Basis |
|---|---|---|
| Using proprietary data to train models | High — infringement claims | Copyright / contract law |
| Claiming AI outputs as owned by developer | Medium — depends on human input | Human authorship requirement |
| Licensing data vs. scraping rights | High | Contract and statutory IP rights |
| Proposition patents on AI‑discovered insights | Low — unless tied to human inventorship | Patent law |
📌 Summary
| Case / Dispute | Key Principle |
|---|---|
| Thaler v. U.S. Supreme Court refusal | AI cannot be recognized as an inventor/author |
| Concord Music Group v. Anthropic | AI training on copyrighted material contested |
| Bartz et al. v. Anthropic | Fair use defense for lawful training data |
| Andersen v. Stability AI | Training without permission may be infringement |
| Li v. Liu (China) | Human user with creative input can own output |
| Anthropic Settlement | Improper training data use can be financially costly |
🧠 Conclusion
Ownership disputes involving predictive AI systems — including global trade‑flow prediction models — are legally grounded in broader IP and data ownership doctrines. While specific trade‑flow case law may be emerging, the principles from existing litigation over AI training, copyright ownership, and data sourcing directly apply. Clear contracts and human contribution documentation are essential defenses, and courts globally are reinforcing the need for human authorship or licensing before granting IP rights over AI outputs.
RELATED Blog
comments