Originality and Authorship: BCIs in the Realm of Intellectual Property

Eleanor Chou

Edited by Samantha Tonini, Braxton Bullock, Judge Baskin, and Sahith Mocharla

Thoughts, creative expression, and personal data have been regarded as inviolable and private, understood to be owned by the individual under an array of legal protections. Therefore, for years, individual thoughts have been considered beyond the scope of intellectual property law. However, emerging advancements in neurotechnology are challenging this notion. One such advancement is seen in the development of brain-computer interfaces (BCIs), a developing technology that collects, analyzes, and relays brain signal data to devices that output desired actions, such as movement in a prosthetic or a responsive modulation to activity [1]. Within this umbrella lies semantic decoding, a communication BCI that is able to translate individual brain activity into text representative of an individual’s imagined thoughts [2]. The output of the device, being an imperfect transcript, sparks the question of ownership in the legal sphere. Who owns the rights to the outputs of the algorithm, and would they belong to the individual the model was trained on, or the model itself?

With the complex output of BCI technology presenting newfound questions, examining parallels in the field of translating creative works provides a precedent for understanding its role in ownership. In Feist Publications, Inc. v. Rural Tel. Serv. Co. (1991), originality was established as the sine qua non of copyright, which, by definition, constituted a work made independently by the author with a baseline degree of creativity applied, such as in adaptations of plays [3]. In this case, Rural Telephone had collected the addresses, numbers, and names of its subscribers in order to create a telephone book in northwest Kansas. The defendant, publishing company Feist Publications Inc., sought Rural Telephone’s listings for an interregional white-pages directory. When Rural Telephone refused to license its telephone book to Feist, they copied the telephone book’s listings directly, leading to Rural suing them for copyright infringement. While the lower court ruled in favor of Rural, deeming the effort they put into their listings as sufficient for copyright protection, the Supreme Court reversed their decision, establishing a baseline originality requirement for copyrights. Here, they detail a modicum of creativity in original works, expressing that any work with a “minimal creative spark” produced with any degree of intellectual effort qualifies as a creative work. Though this requirement nods to intellectual effort, this is not to be confused with the time put into a work. The court rejects the “sweat of the brow” doctrine, as work cannot be deemed original and independent through sheer effort in creation by the author. In conjunction with the originality requirement, the translated work must exceed a baseline length to be considered for copyright protection [4]. As decided in Signo Trading Intern. Ltd. v. Gordon (1981), short phrases and singular words may not be considered for ownership protections due to their mechanical nature under 17 U.S. Code § 102(b), where copyright laws do not cover systems, methods, and rules. 

While laws governing the protection of translated works are well defined, newer technologies, such as BCIs, pose a new challenge to the definitions of originality and ownership. Through increasingly sophisticated BCI technology, extensive neural data is becoming more directly drawn from individuals. Invasive electrophysiological data collection methods, like neuropixel recordings, can capture focused neural signals with high accuracy. Additionally, computer algorithms are being developed in conjunction with an individual’s neural patterns. Such is the case in semantic decoding, where fMRI data is coupled with a language decoder model that calculates the probability of a given word being thought. To collect the data in this project, three individuals each listened to 16 hours of spoken material as fMRI technology recorded changes in the blood-oxygen levels of their brains. This data was then processed in an algorithm that generated likely word sequences and predicted matching brain patterns, creating a model that could decode the thoughts of the individual whose data the algorithm was trained on. Since the model is predictive in nature, the extracted semantic features are distinct from the individual’s true intended speech. Due to variations in the data extracted from the individual, the calculations responsible for assigning likelihood to words, and additional adjustments through model optimization, the BCI outputs an adaptation of the individual’s thoughts. If the output of the BCI constitutes such an adaptation, copyright precedents are essential in navigating the complex legal debates that follow. 

Due to heavy restrictions on machine ownership, copyright law would likely name the individual whose neural data fuels the semantic decoding as the primary owner.  Currently, computer algorithms are not subject to ownership. Existing regulations hold that works generated by non-humans cannot claim protections [5]. More relevantly, the courts have rejected attempts to copyright AI-generated works as established in Thaler v. Perlmutter (2025), which affirmed human authorship as a “bedrock requirement” of copyright [6]. In Thaler, the plaintiff attempted to register an AI-generated artwork—named “A Recent Entrance to Paradise”—with the U.S. Copyright Office. The plaintiff would then sue after the office rejected his submission, and the district and D.C. Circuit courts would later affirm the denial of his work. In these decisions, the courts cited the Copyright Act’s detailed provisions, which assume human authors. Non-human authors, such as AI models, are unable to provide signatures, information for lifespans, and heirs to calculate copyright. Furthermore, under the Copyright Clause of the U.S. Constitution, the primary purpose of copyright law is to promote creative expression in science and the arts [7]. Since AI cannot be incentivized by social or economic gain, extending copyright protections to respective algorithms would be futile.

The precedent details that expression requires human authorship, intent, and human originality. Regardless of the sophistication of semantic decoding and its newfound ability to mimic complex human processes, its output lacks the required consciousness to establish protections. Though intricate algorithms introduce a degree of creativity that may fall under the adaptation umbrella in copyright, the nature of semantic decoding remains a work derived from the source. In the case of semantic decoding with BCI technology, the differences in output and originating thought stem from differences in semantics rather than diverging intents. Therefore, while the humans behind the decoding algorithm may claim protection for their software creations, the actual computer-generated predictive analysis, as discussed in the case of semantic decoding, falls under systems and methods that are not applicable for copyright [8]. 

---

[1] Jerry J. Shih et al., Brain-Computer Interfaces in Medicine, National Library of Medicine (2012), https://pmc.ncbi.nlm.nih.gov/articles/PMC3497935/.

[2] Jerry Tang et al., Semantic reconstruction of continuous language from non-invasive brain recordings, 26 Nature Neuroscience 858-866 (2023), https://www.nature.com/articles/s41593-023-01304-9. 

[3] Feist Publications, Inc. v. Rural Tel. Serv. Co., 499 U.S. 340 (1991).

[4] Signo Trading Intern. Ltd. v. Gordon, 535 F. Supp. 362 (N.D. Cal. 1981).

[5] Naruto v. Slater, No. 16-15469 (9th Cir. 2018).

[6] Thaler v. Perlmutter, No. 23-5233 (DC. Cir. 2025).

[7] U.S. Const. art. 1, § 8, cl. 8.

[8] See [4].