IP Regulation Of AI-Curated University Admissions Scoring Models.
1. Introduction: AI in University Admissions
Universities increasingly use AI-based scoring models to assess applicants. These models analyze:
Academic records (grades, test scores)
Essays and personal statements
Extracurricular activities
Recommendation letters
Behavioral or demographic patterns
From an IP perspective, issues arise around:
Ownership of AI models – Who owns the algorithm? The AI developer or the university?
Copyright – Does the software, its training data, or output qualify for copyright protection?
Trade secrets – Are scoring algorithms proprietary?
Patents – Can a novel AI scoring algorithm be patented?
Data rights & privacy – AI models use sensitive personal data.
2. IP Issues in AI Admissions Models
a. Copyright
The code of the AI model is copyrightable.
Raw outputs (like applicant scores) are facts, which cannot be copyrighted.
Using copyrighted essays, test prep content, or datasets without permission may infringe copyright.
b. Trade Secrets
AI algorithms can be protected if:
They are secret
Provide economic value
Are reasonably protected by the owner
Reverse engineering or leaks can lead to legal action.
c. Patentability
AI scoring models may be patentable if they are novel, non-obvious, and technically inventive.
Mere weighted scoring or predictive ranking is often considered abstract and non-patentable.
3. Key Case Laws Related to AI and IP
Case 1: Alice Corp. v. CLS Bank International, 573 U.S. 208 (2014) – Software Patent Eligibility
Facts: Alice Corp claimed patents on a computer-implemented method to mitigate financial settlement risk.
Ruling: Patents covering abstract ideas implemented on a computer are invalid unless they include an inventive concept.
Implication for AI Admissions:
A model that only sums weighted grades and test scores is too abstract to be patented.
Patent claims must involve technical innovation, such as a unique method for predicting student success using AI.
Case 2: Authors Guild v. Google, 804 F.3d 202 (2nd Cir. 2015) – Copyright & Data Use
Facts: Google scanned books for indexing; authors sued for infringement.
Ruling: Transformative use for search indexing qualifies as fair use.
Implication:
AI admissions models using essays or test prep content may not infringe copyright if the use is transformative (e.g., for scoring, analytics, or predictive modeling).
Case 3: Diamond v. Diehr, 450 U.S. 175 (1981) – Software Patentability
Facts: Patent on a computer-controlled rubber-curing process.
Ruling: Software can be patentable if applied to a technical process.
Implication:
An AI admissions model may qualify for patent protection if it solves a technical problem in predictive analytics, rather than just ranking scores.
Case 4: Waymo v. Uber (2017) – Trade Secrets
Facts: Waymo sued Uber for misappropriation of self-driving car trade secrets.
Ruling: Misappropriation of confidential algorithms and code is actionable.
Implication:
Universities must treat AI scoring algorithms as trade secrets to protect against leaks or reverse engineering.
Case 5: Oracle America, Inc. v. Google LLC, 593 U.S. ___ (2021) – API & Copyright
Facts: Google copied Java APIs to build Android; Oracle sued.
Ruling: Transformative use of APIs may fall under fair use.
Implication:
AI admissions software often relies on third-party libraries; proper licensing is critical to avoid infringement.
Case 6: SAS Institute Inc. v. World Programming Ltd., 2013 (UK/EU) – Software Functionality
Facts: SAS sued WPL for replicating software functionality without copying code.
Ruling: Software functionality cannot be copyrighted, only the source code.
Implication:
Admissions scoring logic (weighting GPA, test scores, essays) cannot be copyrighted, but the code implementing it can.
Case 7: Thaler v. Commissioner of Patents (Australia, 2021) – AI as Inventor
Facts: Stephen Thaler applied for patents listing an AI system as inventor.
Ruling: Courts ruled only humans can be inventors under current law.
Implication:
AI-generated admissions models cannot be patented if the AI alone “invented” the method.
A human developer must be listed as the inventor.
Case 8: IBM v. Priceline (Trade Secret/AI Prediction)
Facts: IBM claimed trade secret protection over predictive AI models used in pricing.
Ruling: Courts reinforced that AI models with confidential data and methodology are trade secrets if adequately protected.
Implication:
Admissions scoring models using sensitive student data qualify as valuable trade secrets.
4. Practical Takeaways for Universities
Protect AI Scoring Models as Trade Secrets
Keep source code, training methods, and weighting schemes confidential.
Copyright Compliance
Avoid copying essays, test prep materials, or third-party datasets without permission.
Transformative use (analytics) may reduce infringement risk.
Patent Considerations
Focus on technical innovation, like novel predictive algorithms, not abstract weighting.
Humans must be listed as inventors; AI cannot legally be an inventor.
Licensing Third-Party Software
Ensure libraries or datasets used are properly licensed to avoid copyright issues.
Balancing Transparency and IP Protection
Regulations may require AI explainability for fairness and bias auditing.
Protect proprietary algorithms while ensuring compliance with anti-discrimination laws.
✅ Summary
Trade secret law is the primary tool to protect AI admissions models.
Copyright protects software code, not the output or abstract scoring logic.
Patents are limited to AI methods demonstrating technical innovation.
Fair use doctrines may shield transformative use of training data.
Courts increasingly recognize AI-related IP issues, especially trade secrets, patent eligibility, and copyright of underlying code.
RELATED Blog
comments