Copyright Implications For BrAIn-Signal Datasets And Neuro-Data OwnershIP.
1. Understanding Brain-Signal Datasets and Copyright
Brain-signal datasets (like EEG or fMRI recordings) raise several legal questions:
Raw neural data: The raw voltage fluctuations or MRI images are essentially facts about a person’s neural activity. In most jurisdictions, facts are not copyrightable, though compilations of these facts may be.
Processed or annotated datasets: When datasets are curated, cleaned, labeled, or annotated, they may have sufficient originality to attract copyright protection.
Ownership vs. control: The person whose brain generated the data (the subject) may not automatically own copyright. Usually, ownership depends on contracts, consent forms, or research agreements.
Derivative works: Machine-learning models trained on brain-signal data can raise questions about derivative copyright if the dataset itself is copyrighted.
2. Key Legal Principles
Facts vs. Original Expression: Facts themselves cannot be copyrighted (Feist Publications, Inc. v. Rural Telephone Service Co., 499 U.S. 340 (1991)), but original compilations of facts may qualify.
Data Ownership and Consent: Individuals may retain rights over their own personal data under privacy laws (e.g., GDPR in Europe), but copyright is separate.
Database Rights (EU): In Europe, sui generis database rights protect substantial investments in compiling data.
Neuroscience-Specific Considerations: No court has directly ruled on EEG/fMRI copyright, but analogies can be drawn from datasets in other scientific domains.
3. Illustrative Case Laws
Case 1: Feist Publications, Inc. v. Rural Telephone Service Co., 499 U.S. 340 (1991) – U.S. Supreme Court
Issue: Whether a telephone directory’s compilation of names and numbers could be copyrighted.
Holding: Facts are not copyrightable; only original selections/arrangements are.
Implication for neurodata: Raw EEG/fMRI readings are like facts—they are not copyrightable. Only curated, annotated datasets may qualify if there is originality in selection or presentation.
Case 2: International News Service v. Associated Press, 248 U.S. 215 (1918)
Issue: Misappropriation of news content.
Holding: Courts recognized a quasi-property right in “hot news” for limited time.
Implication for neurodata: If a lab invests significant effort in collecting brain-signal data, they may claim ownership of compiled datasets, though copyright may not apply to raw signals. Analogous to “hot news” doctrine.
Case 3: A.V. v. iParadigms, 562 U.S. 221 (2011)
Issue: Ownership of student submissions in plagiarism detection systems.
Holding: Students retained copyright; the system could store and process for specific purposes under license.
Implication: Neural data collected for research may remain the participant’s “property” in terms of copyright, depending on consent agreements. Researchers often get a license to use the data, not ownership.
Case 4: Bridgeman Art Library v. Corel Corp., 36 F. Supp. 2d 191 (S.D.N.Y. 1999)
Issue: Copyright for exact photographic reproductions of public domain artworks.
Holding: Exact reproductions lack originality; therefore no copyright.
Implication: EEG/fMRI recordings are essentially mechanical or observational reproductions of brain activity. If the dataset is a direct record, copyright may not attach unless transformed creatively (e.g., visualizations or derivative analyses).
Case 5: Kelly v. Arriba Soft Corp., 336 F.3d 811 (9th Cir. 2003)
Issue: Use of copyrighted images in search engine thumbnails.
Holding: Thumbnail use was fair use; court recognized transformative use.
Implication: Transformative processing of brain data (e.g., anonymization, aggregation, feature extraction) could be protected as derivative work, or may be permitted as fair use in research contexts.
Case 6: Oracle America, Inc. v. Google, Inc., 750 F.3d 1339 (Fed. Cir. 2014)
Issue: Use of Java APIs in Android.
Holding: APIs can be copyrighted, but fair use may apply depending on context.
Implication: Structured brain-signal datasets with a specific schema could resemble software APIs. Ownership may extend to dataset structure, but not the underlying signals themselves.
Case 7: Authors Guild v. Google, Inc., 804 F.3d 202 (2nd Cir. 2015)
Issue: Google scanned books for search indexing.
Holding: Transformative use of copyrighted works can be fair use.
Implication: Machine learning or AI analysis of neurodata may be transformative enough to avoid infringing copyright, even if derived from copyrighted compilations.
4. Practical Implications for Neuro-Data Owners
Raw neural signals: Generally not copyrightable.
Curated datasets: May qualify if there is original selection/organization.
Consent & contracts: Essential to define who can use, store, and publish the data.
Derivative works: AI models trained on the data might raise copyright questions.
Cross-jurisdiction differences: U.S. vs EU (Database Directive) approaches differ significantly.
5. Summary Table: Neurodata & Copyright Analogy
| Neurodata Type | Copyright Status | Legal Analogy Case |
|---|---|---|
| Raw EEG/fMRI signals | Not copyrightable | Feist v. Rural Telephone |
| Curated & annotated dataset | Possibly copyrightable | Bridgeman Art Library v. Corel |
| Transformed dataset (AI-ready) | May be derivative work | Authors Guild v. Google |
| Data compilation effort | Ownership via contract | INS v. AP |
| Schema / database structure | May have IP rights | Oracle v. Google |
In short, ownership of brain-signal data is mostly governed by agreements, consent, and privacy laws, while copyright protection depends on originality in curation or transformation. Courts have not directly addressed EEG/fMRI, but these cases provide strong analogical guidance.
RELATED Blog
comments