Should Software Copyright Protect the Whole Program or Just Its Code? — On the Copyright Eligibility of AI-Assisted Code and the Practical Dilemma of Registration

With the deep penetration of generative AI technology in software development, large language models now possess the powerful ability to directly generate high-quality source code, posing a severe challenge to the traditional computer software copyright system. Currently, the China Copyright Protection Center requires applicants for software copyright registration to submit a written pledge guaranteeing that the software contains no AI-generated code whatsoever. This “no AI” pledge requirement has sparked widespread controversy. The core question—whether the primary subject of software copyright protection should be micro-level code characters or the macro-level overall structure of the software—urgently needs clarification. This article traces the legal origins of the subject of software copyright protection, explores the limitations of traditional code-centrism in the intelligent age, and demonstrates the reasonability of protecting software as a whole and its “structure, sequence, and organization.” Drawing on the latest judicial reasoning and core legal theories, this article thoroughly examines the boundary of originality determination for software works under human-machine collaborative development. Code purely generated by algorithms or simple lists of prompts, lacking human individualized expression, can hardly satisfy the originality requirement for works. Yet the overall software developed by humans using intelligent tools, embodying significant intellectual labor in complex architectural planning and logical arrangement, deserves copyright protection. The Copyright Center’s practice of fully rejecting AI-code applications for software copyright, though perhaps intended to prevent excessive monopoly of public domain resources, severely overlooks the core intellectual contribution of humans in overall software architecture and even artificially creates unequal R&D status among developer groups. China’s software copyright registration system urgently needs to transition toward a mechanism of disclosure and exception declarations under formal examination, establishing disclosure obligations for AI-assisted generation and introducing rights reservation declarations to achieve a dynamic balance between encouraging technological innovation and protecting the public interest.

Keywords: Intellectual Property; Software Copyright; Artificial Intelligence; Software; Originality; Examination Practice;

* This article reflects only the author’s personal views and is not intended as legal advice.

Introduction

Since China established its software intellectual property protection system centered on the Copyright Law and the Regulations on the Protection of Computer Software, the traditional path of software protection has always been based on human programmers writing code by hand. In this traditional model, each line of source code typed by an individual developer or a team of programmers was naturally regarded like the words in a literary work, straightforwardly protected by copyright law.

However, the wave of information technology is reshaping the global economy and social structure at an unprecedented pace. Among these changes, the explosion of generative AI has become a core driving force behind a new round of industrial transformation. In the field of software development, human-machine collaborative development has gradually become the industry mainstream. New AI programming tools represented by GitHub Copilot, Claude Code, Codex, Cursor, TRAE, etc., can now directly and intelligently generate hundreds or even thousands of lines of functional code based on developers’ natural language instructions through chain-of-thought reasoning and behavioral observation interaction (the “think-act-observe” paradigm), greatly improving R&D efficiency.

While this technological revolution has enormously boosted software engineering productivity, it has also created a huge impact on existing intellectual property rights confirmation practices. In current administrative practice, software developers face unprecedented strict scrutiny when applying to the China Copyright Protection Center for computer software copyright registration. The examining authority explicitly requires applicants to make an absolute pledge that their software has not used AI to generate any code. If applicants truthfully disclose that the software contains algorithm-generated code segments, they are directly refused registration.

This zero-tolerance administrative practice has effectively touched upon the core issue at the bottom of the software copyright system. Is the subject protected by copyright law the code instruction sequence itself at the microscopic physical level, or the overall software structure and business logic constructed by these codes? If the focus of protection falls entirely on code characters, then machine-generated code clearly cannot satisfy the legal requirements for authorship. If the focus shifts to the non-literal expression of the software as a whole, the enormous intellectual investment by human developers in architectural design, function definition, prompt engineering, and later-stage system debugging should be sufficient to bestow copyrightability upon this software system containing machine code. This article intends to conduct a systematic argument on this issue from the dual dimensions of legal analysis and judicial/administrative practice.

1. The Legal Origins and Evolution of the Subject of Software Copyright Protection

To explore the eligibility of intelligently generated code, we must first clarify at the source exactly what kind of intellectual achievement software copyright protects. For a long time, both theoretical and practical circles have held differing understandings, primarily forming two distinct views: “code centrism” and “software holism.”

1. Traditional “Code Centrism”

The current legal framework initially tended to treat computer software similarly to literary works protected by copyright law, adopting a code-centrist protection path highly dependent on physical carriers. Article 2 of China’s Regulations on the Protection of Computer Software defines computer software as “computer programs and their related documentation,” while Article 3(1) explicitly defines a computer program as “a coded instruction sequence that can be executed by a computer or other information processing device to obtain a certain result, or a symbolic instruction sequence or symbolic statement sequence that can be automatically converted into a coded instruction sequence.” This legislative formulation directly places “code” at the absolute core of software definition. Therefore, in previous software copyright registration practice, submitting 30 consecutive pages of source code before and after became the core examination material. The underlying logic of this operational model treats source code as equivalent to the literary expression in written works, believing it directly reflects the creator’s intellectual choices and arrangement, forming what is known as the “code-centrist” protection view.

In the early stages of software development, this view indeed had historical reasonability. At that time, programming languages were relatively low-level, development tools were primitive, and programmers had to invest substantial intellectual labor to write code line by line. For similar software functionality requirements, different programmers would create vastly different concrete code texts based on their individual understanding. Their personal coding style, logical ingenuity, and even extreme algorithm optimization skills were clearly reflected in the code text. At that point, protecting literal code was almost equivalent to protecting the software’s core value.

Consider the US case Computer Associates International, Inc. v. Altai, Inc. from the 1990s. Prior to this case, judicial practice had shown that the literal components of program code were protected by copyright. The dispute in this case centered on whether and to what extent copyright protects the non-literal components of program code—the structure, sequence, and organization (SSO) of the program.

In this case, plaintiff CA Company discovered in 1988 that defendant Altai was using code from its CA-ADAPTER software and sued Altai for copyright infringement and trade secret misappropriation. In response, Altai had eight programmers who had no knowledge of or involvement in the original OSCAR code rewrite the copied code portions in a completely confidential environment (using entirely different code to achieve the same functionality) and released the new program as OSCAR 3.5.

The U.S. Court of Appeals for the Second Circuit, after filtering out all functional and public domain elements, compared the two codes and found that Altai “after rewriting OSCAR 3.4 into OSCAR 3.5, had almost no code lines identical to ADAPTER,” with only a few lists and macros similar to ADAPTER that did not sufficiently impact the program to constitute copyright infringement. Additionally, the court held that similarities in services required by the operating system stemmed from the operating system’s own characteristics and were not copyrightable; likewise, flowcharts were deemed influenced by external factors determined by the nature of the work and were similarly unprotected. Based on this analysis, the court affirmed the district court’s ruling that OSCAR 3.5 did not constitute copyright infringement, confirming early judicial emphasis on literal code expression at the underlying level. At the same time, the case established the famous “Abstraction-Filtration-Comparison” three-step test, demonstrating that copyright protection for computer programs is not limited to literal code but also extends to non-literal program elements, including overall architecture, module division logic, and data structure organization.

2. Technological Iteration Diminishes the Originality of Micro-Level Code

With continuous iteration of programming technology, the proliferation of high-level programming languages and modern software engineering exhibit characteristics of high modularization, framework-orientation, and open-source adoption. Particularly with the application of AI programming tools mentioned earlier, the foundation of “code centrism” has been shaken. Its limitations are mainly reflected in two aspects.

First, the space for “original expression” at the code level has been significantly compressed. Copyright law contains the classic “idea-expression dichotomy” principle: copyright only protects “expression,” not “ideas.” When the expression of a particular idea is extremely limited, or a certain expression becomes the only or optimal way to implement that idea, the expression merges with the idea and loses eligibility for copyright protection—a phenomenon known as the “merger doctrine.” As software technology develops and programmers continuously exhaustively optimize underlying techniques, many standardized functional functions have limited “optimal writing methods.” For basic functions such as general data sorting and routine interface calls, code across different projects very easily present high homogeneity, degenerating from original expression to general-purpose tools in the public domain. AI programming models, by learning massive open-source code libraries and undergoing extensive “human preference fine-tuning,” tend to generate optimal or standardized solutions for specific functional requirements. At this point, clinging to literal code similarity comparison would completely deviate from copyright law’s legislative intent to encourage innovation rather than monopolize common knowledge.

Second, this view cannot cover the core intellectual activities of modern software development. As development barriers lower, the commercial value of modern software engineering has shifted from the early goal of “ensuring the program runs” to front-end activities: requirements analysis, system architecture design, module division, data structure definition, and UX (user experience) design. These high-level intellectual activities determine the software’s functionality and performance, and even whether the software can successfully reach the market and gain user acceptance. Yet their value cannot be fully reflected in specific micro-level code segments. If we limit our view solely to the ultimately written coded instruction sequence text, we may violate the objective laws of modern software development.

3. The Legal Justification and Practical Need for Software Holism

Facing the objective decline of micro-level code originality and the high replicability of functionally identical code, Chinese courts have explicitly recognized the legal value of non-literal software expression in numerous software infringement disputes, affirming that the scope of software copyright protection should extend to the overall software structure.

For example, in the “Tai Chi Panda vs. Hua Qian Gu” game copyright infringement case, the court found that Hua Qian Gu had engaged in “skin-swapping” plagiarism of Tai Chi Panda and utilized specific expressions of Tai Chi Panda’s gameplay rules, constituting copyright infringement. The judgment explicitly stated that “the interface layout and content of this part can be regarded as a specific presentation of the game’s specific gameplay rules, constituting ‘expression’ under copyright law.” Although this case treated the entire game graphics as an “audiovisual work” for copyright infringement determination, it equally proves that commercial software is not merely composed of underlying code—other elements such as art resources and interactive rules are also protected by copyright law.

Regarding the “specific gameplay rules” at issue, in actual commercial development, they are often written by game designers in documents or spreadsheets. The code’s function is merely to digitally present these rules. Without the underlying code, the rules text cannot be presented on the terminal; equally, without these rule texts, code loses any execution logic. Within the engineering system, neither is superior.

Furthermore, in another case, the court’s examination path further confirmed the independent legal status of non-literal expression. After comparison, the court confirmed that the source code languages and compiled languages of the two software programs were completely different. Despite the underlying literal code having no connection, the court still initiated a substantive similarity review of the two programs’ interactive interfaces and functional design layers. Although the court ultimately found that the similar interfaces and functions were customary designs in that service industry software and lacked originality, thus constituting no infringement, this judicial logic clearly demonstrates that judicial adjudication has long broken through the limitation of single code comparison, establishing a judicial consensus that software business architecture and functional logic can be independently considered as subjects.

When developers use intelligent tools for assisted programming, although the underlying micro-level code is directly generated by AI, the software’s entire directory hierarchy, module interaction flow, core business logic, etc., are essentially entirely conceived and constructed by human developers. Through extremely detailed prompt design, multi-round contextual prompting, and strong manual intervention, developers complete the architectural design, fix errors and bugs in AI-generated code during development, and iterate and optimize the system. The software produced through this development process is essentially full of human wisdom. What copyright law urgently needs to protect is the software as a complete logical system and engineering achievement—it must not be degraded to an isolated judgment of the physical source of individual code segments.

2. Dialectics of Originality of AI-Generated Content from a Judicial Perspective

Having established the legal basis for tilting the protected subject toward the software as a whole, it is necessary to further analyze the copyright law attributes of intelligently generated content itself. Several landmark judicial rulings in China in recent years provide highly instructive guidance for defining the originality of such content.

1. The Judicial Logic of Continuous Intellectual Input Establishing Human Authorship

In the 2023 case before the Beijing Internet Court—China’s first AI-generated image copyright infringement case—the judicial authorities explicitly recognized the substantive intellectual input of human developers when using intelligent tools. The court’s reasoning held that when users invest continuous and specific intellectual labor in prompt design, parameter adjustment, and iterative selection of image elements, this behavior has transcended simple mechanical command operation and constitutes original expression under copyright law.

Extending this reasoning to the programming field presents a highly similar picture. When programmers or developers operate large language models to write code, they inevitably expend significant effort on business decomposition and deep prompt engineering to constrain and guide the model to output accurate results. Subsequently, they must assemble, test, fix, optimize, and systematically refactor the generated code modules. These coherent actions equally constitute highly individualized intellectual creation. Thus, the overall software that integrates such extensive human wisdom fully satisfies the core originality requirements for copyright protection.

2. The Idea-Expression Dichotomy and the Denial of Simple Instruction Works

In stark contrast to the above affirmative ruling, the Shanghai Huangpu District People’s Court’s 2025 judgment in China’s first AI prompt copyright ownership case drew a clear legal boundary. In that case, the plaintiff claimed that six sets of prompts covering artistic style, subject elements, and material details constituted protectable literary works. After strict adjudication, the court found that the elements within the prompts were merely simple lists lacking basic grammatical logic and contextual narrative sequence. From a legal characterization perspective, such keyword groups belong to abstract ideas within the realm of thought, failing to demonstrate the author’s individualized intellectual input at the expression level. Based on the idea-expression dichotomy principle, the court legally rejected all the plaintiff’s claims for copyright confirmation.

This judgment clearly conveys another judicial cautious attitude: if simple keyword combinations were easily recognized as works, it would likely restrict the public’s free use of language and vocabulary, thereby hindering the healthy development of the entire intelligent innovation ecosystem. This also inversely confirms that the products of simple instructions lacking complex logical connection cannot obtain copyright protection.

3. AI Works Should Be Recognized Based on the Degree of Human Creative Participation

In contrast to the above cases where humans participated and were credited as authors, in the case of Shenzhen Tencent v. Shanghai Yingxun Technology for copyright infringement and unfair competition, the court found that financial articles automatically written by Tencent’s robot Dreamwriter constituted protected literary works under China’s copyright law. The court’s logic did not wholesale deny copyright simply because the directly generated text came from an algorithmic system; instead, it thoroughly explored the guiding role of humans behind the scenes.

The court explicitly stated in its judgment that the key to determining whether the articles constituted literary works was whether they possessed originality. The judgment process should be conducted step by step. The core standard was to analyze, from the generation process of the articles, whether they reflected the creator’s individualized choices, judgments, and skills. In specifically determining whether the relevant personnel’s actions constituted creative acts under copyright law, it was necessary to consider whether the action was an intellectual activity and whether there was a direct link between the action and the specific expression form of the work. In this case, the creative team invested substantive intellectual labor in data input, trigger condition setting, template design, and logical arrangement. The intelligent program was merely a tool for executing these established logics.

In related commentary, Li Yang, then a professor at Sun Yat-sen University Law School and Vice President of the China Intellectual Property Law Association, publicly stated: “Artificial intelligence is an extension of the human brain and body. Its role in the creative process merely reduces human intellectual and physical labor—no different in essence from the role played by traditional machines in automated industrial production. Since products created by traditional automated machines are considered human labor achievements, treating AI-generated content as human intellectual achievements, i.e., expressions of thought or emotion, presents no legal obstacle.”

Mapping this to software development, when developers use intelligent programming assistants, they similarly play the roles of algorithm rule setter, business logic planner, and final code approver. If the human developer’s participation level remains merely inputting an abstract concept, they naturally cannot claim rights. But if the human developer deeply engages in module decomposition, prompt constraint design, system exception handling logic setting, and code integration and testing, their degree of creative participation has already reached the originality threshold required by copyright law. In this case, recognizing the overall software (including AI-assisted code) as a protected work is precisely the respect and legal recognition of human intellectual labor.

4. The Separation of Rights Between Macro Software Architecture and Micro Generated Code

The three typical cases above collectively illustrate the current judicial status of determining copyright for AI-generated content. Mere micro-level code segments directly generated by algorithms based on simple instructions—for example, a developer using only a single prompt like “please generate a Snake game” for the AI to automatically output an independent program—are like simple lists of prompts lacking logical connection. Because they lack human individualized arrangements, they necessarily fall into the unprotected public domain. But the overall software, which is predominantly planned by humans incorporating complex prompt design and ultimately forms specific commercial functionality, embodying substantive human intellectual achievements and thus possesses overall originality.

On the proposition of whether intelligently generated code is worth protecting, both the legal profession and administrative authorities need to adhere to the “structure, sequence, and organization” logic, allowing reasonable segmented judgment of a software program. Individual functions or regular code segments directly output by algorithms may not enjoy exclusive copyright, but the overall software structure formed under human developer coordination with specific functionality should receive full copyright protection.

3. Practical Issues and Legal Analysis of the Copyright Center’s Full Rejection of Generated Code

Facing the technological wave of AI-assisted programming, the Copyright Center’s “no AI pledge” registration policy reflects the administrative authority’s extreme caution when legal provisions remain unclear. However, this seemingly “safe” approach exposes inherent contradictions and potential obstacles to industry development.

1. Confusing the Functional Boundaries Between Administrative Formal Examination and Judicial Substantive Examination

To discuss the Copyright Protection Center’s administrative decisions, we must first clarify the legal nature of China’s computer software copyright registration system. Unlike patent law, which requires the China National Intellectual Property Administration to conduct strict substantive examination of novelty, inventiveness, and practical applicability, China’s computer software copyright registration is essentially voluntary registration with formal examination. According to the Measures for Computer Software Copyright Registration, the administrative examination authority’s primary responsibility is to verify the formal completeness of submitted application forms, identification materials, and supporting documents. Copyright arises automatically from the date of software development completion; the administrative registration certificate only constitutes prima facie evidence of rights.

In current administrative practice, the Copyright Center has implemented an extremely strict zero-tolerance examination standard for software applications containing AI code. Applicants must submit a legally binding written pledge when filing, explicitly guaranteeing that the software has not used any intelligent system to generate even one line of code. Once examiners discover, or applicants voluntarily disclose, the existence of algorithm-generated code segments, the application faces indefinite shelving or even direct refusal of registration, and the applicant’s dishonest behavior may be recorded in their credit history.

From the administrative authority’s examination perspective, this full-rejection policy has specific risk prevention considerations. China’s current Copyright Law has not yet systematically codified the ownership of rights for generative AI content. The administrative examination department’s highly cautious attitude toward non-human directly created literal code aligns with the administrative principle of “no action without legal authorization.” Fully opening up registration for such code could easily encourage some profit-driven entities to use large models to mass-generate standard code and mass-apply for certificates, thereby maliciously monopolizing public technical resources and even generating numerous frivolous lawsuits for improper gain.

However, the Copyright Center’s actions essentially forcibly relocating the substantive examination authority—determining whether software possesses originality—which belongs to the judiciary, into the administrative registration stage. Simply because the software’s physical composition includes machine-generated micro-level code characters, the administrative authority directly deprives the entire software system’s registration eligibility. This approach lacks solid support in both administrative authorization and legal logic.

2. Zero-Tolerance Requirements Deviate from the Objective Development Law of the Software Industry

Although the administrative intent to prevent rights abuse has some reasonability, the mandatory requirement for applicants to pledge absolutely no generated code has severely deviated from the objective technological evolution trend of the current global software industry.

With the comprehensive proliferation of various intelligent programming assistants, the interaction between human developers and intelligent algorithms has penetrated to the keystroke level of real-time code completion. A human developer need only input a function name and part of the code logic, and AI can accurately predict and complete the subsequent code. The developer simply presses the Tab key to instantly adopt these intelligently generated code segments. In real system development, the boundary between human code and AI code becomes extremely blurred, often presenting a highly integrated state of half-manual, half-AI. Forcing developers to precisely separate every machine-generated character from massive codebases of hundreds of thousands or millions of lines is entirely infeasible from a software engineering technical standpoint.

As discussed earlier, modern software engineering embodies extremely complex systemicity while encompassing rich architectural wisdom. Refusing to issue copyright registration certificates for software systems containing machine code is like depriving an architect of design copyright for an entire building simply because some walls were built using machine 3D printing technology instead of manual pouring. This analogy intuitively reflects the dual impropriety of the zero-tolerance policy in both legal principle and common sense.

3. Stringent Examination Triggers Integrity Crisis and Enforcement Difficulties

An even more severe reality is that this harsh absolute prohibition policy very easily breeds a vicious cycle of “bad money driving out good” in practice. Developers who honestly declare their use of technical aids are rejected at the door, while applicants who choose to conceal their actual practices pass formal examination and obtain legal certificates. This severely undermines the credibility of the national software copyright registration system, leaving the copyright administration unable to accurately grasp the real data and industry baseline of new technology-assisted development in China’s domestic software industry.

When software enterprises that have genuinely invested heavily in system architecture innovation and complex prompt design suffer market infringement, their inability to obtain basic rights confirmation certificates through normal channels leaves them in a passive and difficult position in subsequent judicial enforcement and commercial transformation, with their legitimate rights and interests facing enormous risk of being wantonly trampled.

4. Artificially Creating “Unequal R&D Status”

A modern commercial software program’s components are extremely rich. Beyond the bottom-level execution code, it equally cannot be separated from massive art interface materials, interactive animations, and various embedded text content. Yet the Copyright Center’s current “no AI pledge” registration policy only targets the single dimension of “code.” This means a programmer with pure coding ability but zero art skills could use AI painting tools to batch-generate all art materials and UI for a game or software. As long as the underlying code is manually typed, the software can still smoothly obtain software copyright registration. Conversely, a non-programming developer with high aesthetic sense, business coordination, and prompt logic planning skills, who tries to use AI programming assistants to help generate underlying code while supplementing it with their own original art materials and business logic to build innovative software, would be strictly rejected by the Copyright Center simply because it contains AI-generated code segments.

Currently, China’s digital economy is vigorously promoting the development of “one-person companies” or super-individual models. The fundamental driving force is precisely that the qualitative change in AI technology has finally allowed many people who lack traditional code-writing ability but possess strong business insight and design talent to successfully land their innovative ideas as runnable software products.

The core commercial and social value of software lies in solving practical problems and meeting user needs. If the administrative authority insists on “code-only evaluation” in rights confirmation, granting developers who understand code the privilege of using AI art while depriving designers who understand design the right to use AI code, this essentially establishes “code hegemony,” completely deviating from copyright law’s legislative intent to protect diverse intellectual achievements. This policy direction not only negates the wisdom and efforts of product managers, UI designers, narrative planners, and other R&D participants, but also artificially creates technical barriers to cross-border innovation in China’s software industry.

4. Exploring Paths to Reconstruct Software Copyright Examination and Protection Mechanisms in the New Technology Era

Facing the irreversible wave of technological innovation, clinging to traditional literal code comparison examination methods has shown an obvious era lag. China’s computer software copyright system must undergo modern system reconstruction, shifting the protection focus from mere code character comparison to comprehensively evaluating human intellectual contributions to the overall software logic and system architecture.

1. Establish a Mandatory Disclosure Mechanism Based on Formal Examination

China’s copyright examination authorities should transform their inherent administrative mindset and abandon the rigid policy of requiring applicants to pledge no use of intelligent tools. Based on the formal examination nature of China’s software copyright registration system, requiring examiners to conduct substantive assessment of macro architecture and interface originality for millions of annual software copyright applications is not feasible. However, a mandatory disclosure obligation at the application stage could be comprehensively introduced.

The US Copyright Office, when dealing with registration of works containing AI-generated content, has explored a highly instructive regulatory path. In the review of the AI-involved graphic novel Zarya of the Dawn, the US Copyright Office first canceled the author’s copyright registration. Upon re-review, it clearly established the “disclaimer registration principle”: the human author holds copyright for the overall compilation and text portions of the work but must truthfully disclose and explicitly exclude algorithm-generated independent image elements in the application.

China’s copyright registration authorities could consider learning from this experience. During the rights confirmation stage of software copyright registration, they should issue clear regulations requiring applicants, when submitting the software copyright registration form, to have the obligation to truthfully disclose whether and to what extent intelligent code generation tools were used. This would compel applicants to honestly explain the specific names, application stages, and approximate scope of relevant intelligent tools used.

2. Introduce a Classified Registration and Rights Reservation Declaration System

Based on implementing mandatory disclosure obligations, the Copyright Protection Center should make institutional adjustments within the existing registration certificate system, adding a specific legal category for software containing “auxiliary generated content” and simultaneously attaching standardized rights exclusion disclaimers in prominent positions on the certificate.

For software applications that truthfully disclose containing algorithm-generated code, as long as the applicant also submits identification materials proving their overall system architecture design and macro-level arrangement, the administrative authority should grant registration. On the ultimately issued Computer Software Copyright Registration Certificate, a standardized exclusion disclaimer must be legally attached in a prominent position, clearly stating that the copyright protection scope of this certificate is strictly limited—the exclusive protection does not extend to the algorithmically automatically generated base code characters themselves, only covering the human-original system architecture, specific data organization forms, and code portions uniquely arranged by humans.

Through this composite administrative regulatory approach of “rights confirmation + disclaimer,” honest developers’ legitimate needs for obtaining overall software rights confirmation certificates can be met, providing legal endorsement for subsequent commercial enforcement while effectively alleviating the administrative authority’s legal concerns about certain entities potentially abusing the registration system to monopolize public underlying code libraries.

3. Build a Coordinated Protection Network of Copyright, Trade Secrets, and Patents

Facing the objective weakening trend of copyright protection at the micro code level, the practical community should actively guide software enterprises to build multi-dimensional, three-dimensional intellectual property coordinated protection systems. For core commercial algorithm logic achieved through highly complex prompt engineering—facing the realistic risk that relevant departments may not recognize underlying code copyright under the current judicial environment—enterprises can decisively shift their enforcement strategy.

Enterprises can rely on strict internal technical encryption measures and confidentiality agreement systems to fully incorporate these core algorithms and prompt sequences into the protection scope of the Anti-Unfair Competition Law and trade secret law. For novel technical solutions and operational methods with outstanding practical application value developed by humans using intelligent tools, relevant departments should strongly encourage developers to actively seek invention patent protection. The core of patent protection has always been the technical solution itself, completely unconstrained by the superficial issue of who typed specific code characters, and can provide more solid and reliable legal barriers for bottom-level software technology innovation in the intelligent era.

5. Conclusion

With the rapid evolution of various assisted programming tools, the era of individuals fighting alone to write basic code is fading. The marginal cost of machines directly outputting standard runnable code is accelerating toward zero. If relevant management authorities still cling to past examination conventions and insist on treating micro-level character sources as an insurmountable red line, the applicability of the entire software copyright system will inevitably face severe shrinkage. The proliferation of tools like large language models in development environments is precisely forcing the entire legal profession to deeply rethink the technical essence of software engineering. Commercially valuable software is fundamentally not equivalent to the simple stacking of code characters. Its real competitiveness has long shifted to the developer’s business decomposition ability and system architecture design.

In summary, the current zero-tolerance pledge policy for software rights confirmation in practice is essentially a passive defensive management measure. Blindly requiring applicants to prove their innocence or even shutting out advanced productive forces offers little long-term benefit for solving public resource monopoly chaos or protecting genuine innovation. From the macro perspective of industrial development, rather than expending large amounts of administrative resources on tracing the physical generation path of code, examination departments should conform to the objective laws of technological development and refocus their rights confirmation examination on the overall software architecture design and business expression. China’s existing intellectual property management system can fully resolve the current practical deadlock by adding mandatory information disclosure rules combined with classified registration procedures carrying exclusion disclaimers. Once the basic logic of distributing rights according to substantive contribution is established, management authorities may effectively identify malicious code-hoarding behavior while practically protecting the legitimate rights and interests of developer groups with different technical backgrounds, ultimately prompting the software intellectual property system to return to the correct track of encouraging innovation.

References

[1] Feng Xiaoqing. “On the Idea-Expression Dichotomy Principle in Intellectual Property Law.” Legal Forum. 2009 Issue 2 (Vol. 24, Serial No. 122).

[2] 982 F.2d 693. 119 A.L.R.Fed. 741, 61 USLW 2434, 37 Fed. R. Evid. Serv. 348. COMPUTER ASSOCIATES INTERNATIONAL, INC., Plaintiff-Appellant-Cross-Appellee, v. ALTAI, INC., Defendant-Appellee-Cross-Appellant. No. 762, Dockets 91-7893, 91-7935. United States Court of Appeals, Second Circuit. Argued Jan. 9, 1992. Decided June 22, 1992. Amended Opinion Dec. 17, 1992.

[3] Jiangsu Provincial Higher People’s Court. Tai Chi Panda vs. Hua Qian Gu Game Copyright Infringement Dispute. (2018) Su Min Zhong No. 1054.

[4] Supreme People’s Court. (2023) Zui Gao Fa Zhi Min Zhong No. 1250 Civil Judgment.

[5] Beijing Internet Court. (2023) Jing 0491 Min Chu No. 11279 Civil Judgment.

[6] Shanghai Higher People’s Court WeChat Public Account. “Is It a Work? Shanghai’s First AI Prompt Copyright Case Adjudicated Today.” (https://mp.weixin.qq.com/s/qjoCLmWtjb7wxZ0FYt_iSA)

[7] Shenzhen Nanshan District People’s Court, Guangdong. “Walking with the Law • Intellectual Property | Nanshan Court Concludes the Nation’s First Case Recognizing AI-Generated Articles as Works.” (https://gw.nscourt.gov.cn/nscourt/wzsy/fydt/fyyw/content/post_1447471.html)

[8] Li Yang. “Court Recognizes AI-Generated Content as Works, Copyright Hold.” National Copyright Administration Official Website. (https://www.ncac.gov.cn/xxfb/yjdt/202001/t20200108_49729.html)

[9] zarya-of-the-dawn. copyright.gov. (https://www.copyright.gov/docs/zarya-of-the-dawn.pdf)