[author]Cao Bo

[content]

Copyright Law Regulation for AI-Assisted Generative Content

Cao Bo

Associate Professor, Kaiyuan Law School, Shanghai Jiao Tong University

Abstract：The institutional framework of copyright law is designed to encourage creation, with a primary focus on maximizing the quantity of work. The evaluation of the quality of works is mainly achieved through market mechanism and judicial adjudication. The emergence of AI-assisted content generation introduces new competition advantages in terms of delineating rights, reducing costs and enhancing dissemination. This further disrupts the existing landscape dominated by human-centered content in the cultural market. Copyright law possesses an internal institutional structure that aims to correct the incentive effect, while also carrying implicit normative value in improving the quality of works through regulatory guidance and judicial evaluation. By considering the causal relationship between ideas and expressions and recognizing the overall positioning of copyright law as a means to address technical chal-lenges, a fundamental approach to regulate AI-assisted content generation is proposed. This approach includes implementing intelligent recognition and classification mechanism for AI-generated content on self-media platform, imposing copyright content filtering obligation on audio-visual content platform, and creating a statutory license system specifically tailored to AI-assisted content generation.

Key words：generative artificial intelligence;AI generative content;content generated autonomously by AI;AI-assisted generative content

1.Formulation of the problem

The continuous development of artificial intelligence has significantly improved the work efficiency of many industries, improving or even replacing human labor. Phenomena such as AI painting, AI composing poems, and AI composing music are eye-catching, and the core question is whether AI-generated content should be protected by copyright. Starting from the consensus premise of the concept and constituent elements of the work, the authorship, and the creative process, different researchers may come to diametrically opposed conclusions. With the update and iteration of generative AI, chatbots such as ChatGPT can generate content comparable to human works, and the related copyright topic has once again stimulated heated discussions in the academic community.

However, while theoretical discussions are in full swing, the legislative level has remained silent on this issue, and countries still adhere to the basic premise that creative acts can only be realized by natural persons. The U.S. Copyright Office's Guide to Copyright Registration: Works Containing AI-Generated Material emphasizes that "copyright protects only the product of human creativity The term 'author' in the U.S. Constitution and the Copyright Act excludes non-humans." The "Position Statement on Artificial Intelligence and Intellectual Property" published by the Max Planck Institute for Innovation and Competition in Germany argues that "the results of the 'autonomous' generation of AI are obviously not protected by copyright". In the case of Filmin v. Baidu Netcom, the court also clearly pointed out that "the completion of the creation by a natural person should still be a necessary condition for a work under the Copyright Law". At the same time, AI-assisted content is generally considered to be a work within the meaning of copyright law. The reason for this is that the creative process is intervened or guided by humans, and AI only plays the role of a tool. The World Intellectual Property Organization's policy paper on AI proposes to distinguish between "AI-generated content" and "AI-assisted content", implying that the latter can obtain copyright protection. It can be seen that it is not realistic to make a one-size-fits-all evaluation and judgment on whether AI-generated content can obtain copyright protection, and it is a more rational choice to explore a refined normative system through the method of type demarcation. As subordinate concepts of AI-generated content, AI-generated content and AI-assisted generated content are the commonly accepted classification methods.

Under the circumstance that AI-generated content cannot obtain copyright protection, AI-assisted generated content has the possibility of obtaining copyright protection, so it has become a more incentivizing rule expectation to support corresponding business practices and claims. In the context of algorithm push and the popularization and application of mobile intelligent terminal devices, AI-assisted generated content has appeared on a large scale and has become a cultural phenomenon that cannot be ignored. Taking the film and television commentary of the short video platform as an example, the three-minute audio-visual fragments have realized the cutting of film and television works, renaming the movie characters with "Xiaomei" and "Xiaoshuai", with unwavering intelligent voice narration and cookie-cutter background music, while quickly arousing the audience's interest, it immediately became a fleeting "electronic mustard" at the fingertips. When AI-assisted content is associated with boring, monotonous and boring, and gradually reveals the characteristics of patterning, process and homogeneity, it is necessary to reflect on whether the work protection incentives provided by the Copyright Law are legitimate and necessary. On this basis, this paper further examines whether the corresponding theoretical basis and institutional tools can be found within the Copyright Law, so as to realize the correction of the incentive effect and reconstruct the regulatory path of the Copyright Law for AI-assisted generated content.

2. The competitive advantage of AI-assisted content generation

Different from the traditional creation tools, the application of artificial intelligence in the creation process has created a difficult problem in the separation between AI-assisted generated content and AI-generated content on the one hand, and on the other hand, it has profoundly changed the creation ecology and brought significant advantages to AI-assisted generated content. First, compared with human assisted behavior, AI-assisted generated content reduces the cost of communication and exchange, and the claim is simpler, which provides convenience for clearly defining the ownership of copyright; second, AI-assisted generated content is based on data application, which has the characteristics of modeling, which significantly reduces the cost of creation; Third, with the help of algorithmic push, AI-assisted generated content is more in line with the communication characteristics of the mobile Internet, and occupies the initiative in the market competition of cultural products.

2.1The problem of dividing AI-assisted generated content from AI-generated content and AI-generated content independently

The premise that AI-assisted generated content can obtain copyright protection is based on the effective identification of AI-generated content. However, in the application scenario of generative AI, it is difficult to identify the two from the outside. In addition, in the case of the application of AI for content generation, there are differences in understanding on a case-by-case basis as to whether the "creation" process is intervened or guided by humans, and whether AI is merely a tool.

In practice, most of the instances of denying the copyright claim of AI-generated content are based on the premise that the corresponding content has been disclosed or proved to be independently generated by AI. The U.S. Copyright Office (USPO) rejected the applicant's claim for registration of the work of the comic on the grounds that the content was generated by the artificial intelligence drawing program Midjourney, rather than being created as a tool. When Taylor, an American, filed a petition with the U.S. Copyright Office for the registration of the work "The Shortest Path to Heaven", he also made it clear that the content was "created" without any human intellectual contribution, and the U.S. Copyright Office rejected the registration application on the grounds that in the absence of creative input from human authors, the content generated by machines or purely mechanical processes is not a work protected by copyright law. In the case of Filmin Law Firm v. Baidu Netcom, the court found that the article in question was generated by artificial intelligence software, and determined that the graphics and text in the corresponding article did not constitute a work within the meaning of the Copyright Law on the basis that the completion of the creation by a natural person should be a necessary condition for the work. However, generative AI already has the "ability to emerge" to a certain extent, and it is difficult to effectively distinguish the content generated through ChatGPT from works created by humans in terms of expression. If the claimant deliberately conceals the fact that it uses AI to independently generate relevant content, it is very likely that the result of providing substantial protection for AI-generated content in the form of AI-assisted generated content is very likely.

The conclusion that AI-assisted generated content can obtain copyright protection is a logical extension of the definition of creative acts in the Copyright Law. From the point of view of the subject, the author must directly determine the expression that constitutes the work, and help other than the creative contribution can only establish an indirect connection with the work; The Regulations for the Implementation of the Copyright Law of the People's Republic of China (hereinafter referred to as the "Regulations for the Implementation of the Copyright Law") exclude the provision of advice and auxiliary work from the act of creation, and the historical experience of photographic works in obtaining copyright protection in various countries reflects this understanding. However, in practice, how to understand the dominance of human behavior in creative activities is controversial, and different interpretive paths will lead to completely different results. For example, in the case of Tencent v. Yingxun, the court held that the generation process of the article in question was led by people, and the Dreamwriter software played an auxiliary tool, so the generated content met the requirements of the copyright law. In a copyright infringement dispute involving AI-generated images, the court found that the corresponding images were works of art on the same grounds that "human dominance", that is, in the process of generating images, the AI software users selected and further added prompts and modified parameters, reflecting the choice and personality judgment of people. In support of the theoretical proposition that ChatGPT-generated content should be protected by copyright, it also emphasizes that AI plays the role of a tool to assist humans in completing creative acts.

Therefore, the distinction between AI-assisted generated content and AI-generated content in a natural sense is obviously in the nature of a factual determination. However, in the real sense, the interpretation of the law and the application of rules are inevitably affected by value judgments, resulting in the following results: AI-generated content is self-generated to obtain copyright protection in the form of AI-assisted generated content.

2.2 Based on the advantages of tool attributes and human dominance

Whether it is the simple perception of dividing and ending disputes, or the macro vision of maximizing social welfare, the definition of rights has always been an important mission of law: "One of the goals of the legal system is to establish a clear definition of rights, on the basis of which rights can be transferred and recombined through the market. "The definition of rights requires a corresponding social cost, which creates the problem of the cost of boundary rights. At the level of law and economics, the consideration of the cost of boundary power focuses on establishing the boundaries of government power and the way in which it is exercised. The cost of boundary rights discussed in the Copyright Law is mainly the communication costs that may occur in the process of confirming the ownership of copyright, as well as the corresponding social costs arising from disputes and conflicts due to potential different claims.

The creation of any work needs to be realized with the help of certain tools, from pen, ink, paper, inkstone to camera, video camera and even computer, these tools are completely controlled by people, and assume the function of transforming and outputting the expression formed by the human brain, and occupy an absolute auxiliary position in the entire creative process, and the definition of copyright ownership will not bring additional social costs. While creating with the help of tools, it is not uncommon to co-create or assist in the creation based on various identities or contractual relationships. The Copyright Law clearly defines the ownership of copyright through the design of the system of statutory attribution of copyright to producers of service works, commissioned works, collaborative works, and film and television works. The most important issue is the recognition of authorship, and only by contributing to the creative content of a work can it be considered "intellectual activity that directly produces literary, artistic, and scientific works", and that the act of creation can be incentivized by copyright protection. However, in practice, there are often controversies about how to distinguish between direct and indirect contributions to the generation of the expression of a work, and how to distinguish between creative acts and auxiliary work, especially the claim of co-authorship. The "Geyue Mountain Martyrs Group Sculpture Case", the "My First Half of Life" Case" and the "Olympic Sports Center Photo Case" are all typical representatives. The negotiation and determination of the authorship may also lead to confusion of the signature and the manipulation of the author's identity on occasions where the status of the subject is very different, such as the relationship between superiors and subordinates, teachers and students, etc., which may lay hidden dangers for subsequent copyright disputes. In the case of the creation of others, these possible infringement disputes can be attributed to the cost of boundary rights due to the communication between subjects due to the ownership of rights and the distribution of benefits.

According to Coase's theory of "relativity of injury", the occurrence of such a cost of boundary rights is almost inevitable, because no matter how the law defines rights, it is difficult to achieve "equal treatment", either allowing A to interfere with B or B to interfere with A. However, in the scenario of AI-assisted creation, this conclusion based on the assumption of rational homo economicus and basically in line with practical experience is difficult to hold. Taking the popular film and television commentary videos on self-media platforms as an example, the auxiliary work that originally needs to be completed by people, such as copywriting, voice entry, picture editing, etc., can be realized through artificial intelligence. The content generated by AI software is not sufficient in itself to constitute a work to be protected, and the software developers behind it do not have the right to obtain copyright protection through content generation. In the process of human-led creation, AI-assisted generated content will not have the claim of co-authors, and it has obvious comparative advantages in terms of the cost of boundary rights.

2.3 Cost advantages based on data application and modeled production

The copyright law sets the rule that creative acts can only be realized by natural persons, which not only reflects the factual logic of creation, but also reflects the value basis for the law to regulate the behavior and interest relationship of people. The limitations of time, brainpower, physical strength, and wealth as an individual determine that works that can be protected by the Copyright Law also have a certain degree of scarcity, which meets the inherent requirements of property rights. Even after the rise of user-generated content (UGC), the phenomenon of "non-professionalization" of creative and disseminating entities has emerged, the creative behavior carried out by people is still limited by costs, which also determines that the incentive targets of copyright law are minority social groups.

Artificial intelligence has been added as an auxiliary tool to creative behavior, which has profoundly changed the existing creative process and significantly reduced the cost of creation. Artificial intelligence for content generation can be divided into two categories: one is to generate textual content through syntax synthesis and learning, and its technical path is "extracting summaries" and "filling templates", which are mainly used in sports news and other fields, and the other is to apply generative adversarial networks (generative adversarial networks) to continuously improve the authenticity of generated samples through the game between generators and discriminators, so as to be applied to artistic creation. Generative AI, represented by ChatGPT, is an upgraded version of generative adversarial networks, which aims to achieve efficient simulation of human probabilistic language. Intellectual property law scholars summarize their content generation process as "intelligent search engine + intelligent text analyzer + intelligent manuscript washer". Regardless of how its technical model is adjusted and changed, the content generation model of artificial intelligence is highly dependent on the acquisition and training of data, and the output results are also patterned. AI developers are also excited to see large-scale use of their software from the perspective of supporting data training, as the content generated by users using AI will be a source of data for continued machine learning.

The characteristics of free use and modeled content production make the application of artificial intelligence to assist content production have a great driving effect. Since many works are composed of different content elements, people can decompose several pieces of content that originally need to be completed by others into specific tasks, and hand them over to different artificial intelligence software to complete, and after simple integration, claim overall copyright for the final content. The production of audio-visual content can be decomposed, and elements such as copywriting, background music, subtitles, pictures, and dubbing can all be realized by artificial intelligence, and its efficiency is greatly improved. In contrast, the production of film and television dramas in the traditional sense is a complex creative process that requires the collaboration of multiple people, the creation and modification of scripts, the selection and creation of music, the organization and editing of directors, and the performance and cooperation of actors, all of which require different natural persons to do it themselves, and the time, brain, physical and capital costs of creation are high. The large-scale application of AI-assisted content generation is very likely to turn the creative process into a completely streamlined operation, greatly reducing the time, brainpower, physical strength and capital invested in creation, reflecting significant cost advantages.

2.4 The advantages of algorithm-based push and mobile Internet communication

The primary purpose of the Copyright Law is to encourage natural persons to create, but the realization of the economic benefits of the work and the establishment and maintenance of the author's reputation need to be achieved through the dissemination of the work. Authors often do not have the funds and skills needed to establish communication media, so the dissemination of works often requires false hands. Even the immediate motivation of the copyright system has been summarized by American scholars as providing a minimum level of security for the disseminators of works to invest in the mass commercialization of their works. After the rise of the Internet, especially the interactive characteristics of Web 2.0, the threshold for the dissemination of works no longer exists, the role of traditional communication media in the commercialization chain of works has been significantly reduced, and the boundaries between authors and readers, creation and dissemination have gradually blurred. However, the interactivity of information dissemination is still highly dependent on online platforms, and influential large-scale online platforms have become new scarce communication media, which play an important role in the communication ecology of works.

While the main media of the dissemination of works have changed from traditional publishers, radio and television stations to the Internet, the status and nature of consumers of cultural products in the communication pattern have also undergone significant changes, and their initiative in the dissemination of works has been greatly improved. The Internet has broken the limitations of time and space, formed a virtual community based on common interests, and further solidified its circle under the influence of social networks, leading to the gathering and differentiation of cultural groups. After the popularization and application of mobile Internet, mobile intelligent terminal devices not only ensure that users are online at any time, but also deeply embed individuals in the ubiquitous information network. The application of algorithmic push further concretely divides the group into individuals, evaluates their personal characteristics through the collection and analysis of personal information and network activity trajectories, and realizes the "hypernuege" that accurately affects individual choices.

Algorithmic push is the result of the joint action of technical means and business logic, and the deep reason behind it is the physical and mental pleasure of personal preferences and mental pleasure being constantly satisfied. Online platforms take advantage of this psychological effect to trap users in an information cocoon but still enjoy continuing to receive their favorite content with addictive technologies such as infinite refresh or scrolling, like buttons, Zeignick loops, extreme personalized pushes, creating scarcity, and interface manipulation. While content acquisition is becoming more and more convenient, the time for personal attention to be placed on a single content is getting shorter and shorter, and the large-scale output of homogeneous content can meet the requirements of algorithm push. The advantages of AI-assisted generation of content in terms of rights and cost can better feedback user needs, more adapt to the technical logic of algorithm push, and have a superior communication effect in the mobile Internet environment compared with works created by humans.

3. The paradox of incentives for AI-assisted content generated by the Copyright Law

The Copyright Law is based on the logical presuppositions and value positioning that encourage natural persons to create, and since it is difficult to retrospectively restore the creative process of human beings, the corresponding system design has formed a complete set of framework systems for formal evaluation in the process of seeking objectivity. This incentive model of "wide in and wide out" has produced an insurmountable paradox in the face of AI-assisted generated content, which has caused a crack in the logical closed loop of copyright law.

3.1 Independently completed presumption rules to encourage the homogenization of AI-assisted generated content

The copyright laws of various countries require that only intellectual achievements that meet the requirements of originality have the possibility of becoming a work, and the Copyright Law of the People's Republic of China (hereinafter referred to as the "Copyright Law") also has the same rules. There is no further explanation of the connotation of originality at the level of legal norms, but it is generally believed that it can be decomposed into two elements: "originality" and "creation", where the former refers to independent creation and originates from oneself. Like the presumption of authorship, the presumption of "independence" applies to the requirement of "independence", unless there is evidence to the contrary that can rebut the presupposition of independent completion. Therefore, the determination of independent creation is basically equivalent to whether or not to plagiarize the prior work of others. If the possibility of plagiarism is ruled out, the requirement of originality is met even if a work is very similar to other prior works. The classic precedents of the U.S. Supreme Court and China's judicial interpretations all recognize this conclusion.

From the perspective of the distinction of creative acts, in addition to independent creation from scratch, a large number of works belong to re-creation on the basis of existing works, and there must be recognizable differences in order to meet the requirement of "independence". Compared with the creation from scratch, the re-creation has a clear comparison object. If the similarity is extremely high, it is presumed that the act is plagiarism or even reproduction, and copying is excluded from the scope of incentives under the Copyright Law. On the premise of satisfying the differences, the determination of the independent completion of the work formed by recreation is still a presumption in essence. The underlying logic that supports the presumption of independent creation is the individuality of people's creative behavior reflected in their works. In the absence of plagiarism, the likelihood of a high degree of similarity between different works is low, which is a fact-verified historical experience. After the author loses his only manuscript, it is difficult to create a work that is completely consistent with the original content, which also shows that the creative process is non-homogeneous, and the creative results will naturally reflect corresponding differences. The presumption rule of independent completion is further embodied in the determination of infringement, and if there is no plagiarism, it is an independent creation, which means that there is no infringement without plagiarism. If the two works are highly similar and the later work is likely to have been exposed to the previous work, it means that there is a high probability of plagiarism, which constitutes infringement. The infringement determination rule of "contact + substantial similarity" is a way of thinking that deduces the subjective from the objective and the process from the result.

In the case of AI-assisted generated content, it is difficult to directly determine whether it is a creation from scratch or a recreation based on an existing work. The content generated by the application of AI is itself based on a corpus collected in the form of data, but it is difficult to establish a direct connection with a specific prior work in the final presented form. This also means that from the level of comparison of individual works with each other, AI-assisted content has obvious differences that can be identified and meets the presumption rule of independent completion. At the same time, due to the consistency of the same AI content generation software in terms of data source and algorithm model, the output content is also highly similar when the user inputs the commands or asks the same questions. However, it is easier for different users to independently use AI-assisted creations to prove that they have not been exposed to each other's works. Therefore, whether it is to prove independent creation from the positive side or to exclude the possibility of infringement from the negative side, the presumption rules of the Copyright Law on independent creation can support the claim that it constitutes a work when AI-assisted generated content forms a highly homogeneous work. This means that the scenario of homogeneous creation, which originally existed only in the theoretical assumption of homogeneous creation to obtain copyright protection when humans dominated the field of creation, will become a reality in the case of the large-scale emergence of AI-assisted generated content.

3.2 Objective criteria for creativity solidify the competitive advantage of AI-assisted generated content

In addition to this requirement to be fulfilled independently, originality also needs to meet the inventive step element. The authorship system emphasizes the personality attributes of the work, and the copyright system does not require the work to reflect the author's personality or personality. Some scholars have analyzed the historical choice of legal transplantation and the institutional design of the copyright law, and believe that the basis of the protection of works in China is the author's rights system, not the copyright system, and whether it meets the requirements of the creativity of the work should naturally be synchronized with the author's rights system, and the work must reflect the spirit, emotion and personality of the natural person. However, "the reflection of personality in the work" may be supported by hindsight interpretation on many occasions, but it is not an inevitable result of the deduction based on the logic of facts, but more of a philosophical or cultural choice. If this understanding is translated into normative provisions, it will lead to the subjectivity and arbitrariness of legal interpretation, because there is a great deal of room for controversy about what constitutes the spirit, emotion and personality of the author, and it is even more difficult to establish the assertion that the work reflects personality after functional works such as graphic works, model works and computer software have entered the field of vision of copyright law.

Judging from the criteria selected in judicial practice and the theoretical propositions of scholars, although the evaluation of creativity often resorts to subjective expressions such as individualized choice and judgment, labor and resources invested, and creative intention, these descriptions are either not recognized due to great controversy, or only play a rhetorical role. When making a specific analysis of whether a work satisfies inventive step, the court always judges from the external expression of the work, and on the basis of comparing the existing work with the usual expression, the objective "distinguishable change" is used as the criterion for judging. In "Qiong Yao v. Yu Zheng", the court's adjudication line clearly reflects this rhetorical argumentation path of starting from objective expression and inferring subjective choice: "The plot selection and ingenious arrangement of the structure of the work and the deductive design of the plot development reflect the author's personalized judgment and trade-offs, and reflect the author's original thinking results. ”

In practice, the evaluation of independent completion and creativity is usually one, with a creation from scratch presumed to be original after excluding the possibility of plagiarism, and a creation based on an existing work is judged on the basis of whether it has a distinctive change. According to the objective criterion of creativity, for AI-assisted generated content, it is possible to prove inventiveness on the basis that it is different from all existing expressions. The Copyright Law lists works by type, and this legislative model is considered to have the role of exemplary guidance, which can provide convenience in finding the law and reduce the burden of interpretation. In the face of a specific expression, it can be intuitively classified into a specific type of work in an external form, without the need to analyze and apply the general concept of the work. Artificial intelligence content generation software itself is based on the collection of typed work material data, and the generated content is also highly consistent with a specific type of work in appearance. Since humans dominate the entire creative process, although the elements of the disassembled work may be generated by different AI software, the distinguishing qualities of their external expression can still support the creative requirement. As a result, AI-assisted generated content can easily meet the objective criteria of creativity and obtain copyright protection, and this incentive effect will further amplify its competitive advantages in terms of boundary rights, cost and dissemination.

3.3 The notice-deletion rule is difficult to solve the large-scale dissemination of AI-assisted generated content

Copyright has the attribute of private rights, and the enjoyment and realization of private rights fully implement the principle of private law autonomy, which "emphasizes the realization of private interests by private individuals through private law autonomy, fully embodies the freedom of legal acts enjoyed by private law subjects within the scope of law, and respects the creation, modification and elimination of private law relations by private individuals by virtue of their own expressions of will or legal acts". Due to special historical and cultural reasons and the concept of private law as a whole, the Copyright Law reflects strong traces of government control, which affects the realization of the corresponding institutional functions. This also shows that the coordination of private law autonomy and government regulation in the reform of copyright law should be subtractive rather than additional. Specifically, in the field of infringement remedies, the nature of private rights means that it is up to the copyright owner to decide whether and how to seek relief for infringement. Only when the corresponding infringement harms the public interest will the public authority take the initiative to intervene. Therefore, Article 52 of the Copyright Law comprehensively enumerates the specific circumstances that constitute civil infringement, while Article 53 lists the acts that may harm the public interest and thus constitute administrative violations or crimes, mainly reflecting the protection of the property rights of the author, which regulates the direct commercial exploitation of the original work without permission, while the act of plagiarizing the work usually only bears civil liability.

In traditional society, the dissemination of works relies on materialized carriers and institutionalized media organizations, and it is difficult for individual use and dissemination behaviors to produce market substitution effects, which are usually not within the scope of attention of copyright owners. The application of digital technology and the popularization of network services have made individuals become independent communication nodes, free sharing has become a culture, and the dissemination of works has lost control. Although network service providers have the ability to intervene technically and adjust the behavior of network users after the fact, the profit-making model of free content and advertising, as well as the private nature of copyright, make them lack the internal motivation to actively intervene in infringement remedies. This factual logic has led to the gradual abandonment of copyright owners in the process of large-scale digitalization and the search for legislative regulations against network service providers, and the conflict of interests between the traditional copyright industry and the Internet industry has become increasingly acute. It is against this backdrop that the U.S. Digital Millennium Copyright Act created a notice-takedown rule with a "safe harbor" attribute to establish exemptions for network service providers and balance their interests. After the introduction of this rule in the field of copyright, China has deviated from the original nature of the exemption clause through the expansion of other civil laws, making it a typical liability clause. In the context of mobile Internet, the role of network service providers in content dissemination has become more prominent, and the notice-takedown rule has become an important institutional tool for copyright owners to protect their rights.

Since AI-assisted content generation is based on a large number of corpora for the reorganization of content elements, even if it gives people a sense of déjà vu from the appearance, it is difficult to directly lock on to a specific work, and it is impossible to easily conclude plagiarism. Even if the copyright owner can prove the existence of substantial similarity under certain circumstances, the cost and dissemination advantages of AI-assisted generated content can easily achieve scale effects, and the continuous generation of content suspected of infringement makes it impractical to claim copyright protection in the form of private remedies. Taking the commentary of film and television dramas that currently exists in a large number of short video platforms as an example, most of them are AI-assisted generated content. Even if there is an infringement, the copyright owner still needs to issue a notice to initiate the notice-takedown procedure. If the content is a secondary creation, it may also constitute fair use, and the alleged infringer may terminate the notice-takedown procedure by sending a counter-notice. In addition, even if the copyright owner is able to successfully use the notice-takedown procedure to temporarily remove, block, or disconnect the infringing content, similar AI-assisted content will continue to appear. For some works that were created earlier but are still within the protection period, after the transfer and licensing of their copyright property rights have been transferred and licensed, the right holder's initiative to protect rights is not sufficient, and the practical effect of the notice-takedown rule is further reduced. The notification-deletion rule, which is based on the attribute of private rights and private autonomy, cannot form an effective constraint in the face of the phenomenon of "intelligent plagiarism" and large-scale dissemination of AI-assisted generated content.

4. The internal institutional structure of the corrective incentive effect of the Copyright Law

The value presupposition of the Copyright Law to encourage the creation of natural persons is based on the premise that human-created content dominates the cultural market. Limited by the characteristics of institutional costs and creative behaviors, conditional incentives in the normative sense have formed a general incentive effect in practice, but the market mechanism can make up for the shortcomings of insufficient incentives for the quality of works by legal norms to a large extent. However, unlike the existing auxiliary creation tools, the creation efficiency of AI-assisted generated content has been unprecedentedly improved, and its homogeneous, modeled and large-scale characteristics are in line with the communication mode of the mobile Internet, and have significant advantages in the competition with human-created content. The effect of the general incentives of the Copyright Law will further strengthen the competitive advantage of AI-assisted generated content, and challenge the premise that human-created content dominates the cultural market. The incentive function of the market mechanism on the quality of works is also difficult to play an effective role in the context of the widespread application of algorithmic push, and the primary legislative purpose of the Copyright Law to encourage the creation of natural persons is facing a dilemma that cannot be realized. Therefore, it is necessary to explore the internal institutional structure of the corrective incentive effect of the Copyright Law, and activate its normative value to guide the improvement of the quality of works.

4.1 The implicit normative value of the Copyright Law to encourage the improvement of the quality of works

As the basis for the legitimacy of copyright accepted by legislation, the incentive theory is the embodiment of the dominant position of utilitarian theory in the field of intellectual property. John Mill believed that human beings are more capable of attaining "higher happiness" from the mind, emotion, and intellect than animals that can only enjoy "primary happiness", and that spiritual happiness is more worthy of pursuit and more valuable than physical happiness. Although there are some doubts in the theoretical community about "higher happiness", culture and science undoubtedly fall under the category of higher happiness. In order to achieve the goal of promoting cultural and scientific progress, it is necessary to maximize high-level happiness at both the quantitative and qualitative levels, and copyright law naturally needs to assume such a function. Therefore, the copyright law will not intervene in the evaluation of the quality of works in a general sense, so that more forms of expression have the opportunity to obtain copyright incentives to achieve quantitative goals. However, many institutional designs in the Copyright Law reflect the role of regulating and guiding the quality of works, and judicial adjudication does not shy away from the direct evaluation of the quality of works, reflecting the implicit normative value of encouraging the improvement of the quality of works.

The requirements of the Copyright Law for the originality of works, as well as the relevant systems around the construction of originality, all embody the concept of guiding and encouraging high-quality creation, and imply the consensus of human values that high-level happiness has an intrinsic rank. For example, the object of protection of neighboring rights is a specific cultural achievement that does not constitute a work, which may also reflect a certain amount of intellectual input. However, the performance of the work is obviously inferior to the creation of the work itself, and the content of the performer's right is naturally weaker than the copyright. In the protection of works of applied art, the guiding function of the quality of the work is more strongly reflected. Recognizing the nature of works of applied art means that they may obtain additional incentives for copyright protection in addition to patent protection, and it is difficult to make a trade-off based on the objective criteria of originality alone, and countries generally adopt the "separability principle" of practicality and artistry. The U.S. Copyright Act states: "A design of a utilitarian article may be considered a work of painting, figure, or sculpture only if it contains features that are distinguishable from and independent of the utilitarian function of the article." "Although China's copyright law does not explicitly stipulate works of applied art, it provides protection for foreign works of applied art through the Provisions on the Implementation of International Copyright Treaties. In infringement disputes, a work of applied art is usually interpreted as a special kind of work of art, and the separability between artistic beauty and practical function must be separable in order to be protected, and the corresponding artistic beauty must reach the height of the work of art.

The specific provisions of the Copyright Law on damages and the application of the law also clearly reflect the direct judgment of the quality of the work. For example, the Copyright Law stipulates that if the actual losses of the right holder, the illegal gains of the infringer, and the royalties of the rights are difficult to calculate, the court may award compensation of not less than RMB 500 but not more than RMB 5 million according to the circumstances of the infringement. Courts in Beijing and Shanghai regard the value of copyright rights as one of the important factors in determining the amount of compensation in local judicial documents, and factors that can reflect the quality of the work, such as "the type of work, the degree of originality, the investment in creation, the difficulty of creation, the creation cycle, popularity, market value, and awards", can be used to measure the value of copyright. This fully shows that the court has corrected the incentive effect of the Copyright Law by judging the quality of the work in the judgment of individual cases, and by judging the amount of damages. American scholars have put forward the idea that originality should be divided into three levels, and works at different levels should be treated differently in terms of the content of rights and legal remedies. Although this idea has not been accepted by legislation, the "transformative use" and "thin copyright protection" created and applied by U.S. courts are considered useful attempts to distinguish between high and low quality works in order to correct the incentive effect. This shows that "the maximization of social welfare is indeed achieved by more creative works, and the copyright system should also provide greater incentives for such works".

4.2 The structure of causality between thought and expression

As a basic principle in the Copyright Law, the dichotomy between thought and expression has the basic function of defining the scope of protection of the Copyright Law. Specifically, the first is to exclude the integrity of certain specific subject forms from the scope of the subject matter of copyright. The second is to determine which constituent elements of the work should be protected by copyright in infringement cases. For example, under the U.S. Copyright Act, "ideas, procedures, processes, systems, methods of operation, concepts, principles, or discoveries are not protected by copyright law, regardless of the form in which they are described, interpreted, demonstrated, or embodied in a work." This means that neither these objects can be protected under copyright law if they appear as a whole or are embodied in the work as constituent elements. Although the Copyright Law does not stipulate this principle, it embodies the distinction between thought and expression in the Regulations on the Protection of Computer Software, and it is generally used as the basis for adjudication in judicial practice. For example, in the Guiding Cases of the Supreme People's Court, the court clearly stated that "what the Copyright Law protects is the original expression in a work".

The dichotomy between thought and expression originates from the argumentation needs of judicial adjudication, and on the basis of the continuous accumulation of case materials, different objects are successively included in the category of thought and expression, making it gradually become an important analytical framework, "simplifying the number of legal relationships and eliminating the cumbersome ethical argumentation and value judgment". This principle is also more often applied in infringement cases to determine the extent to which the elements of the original work should be protected by copyright. The consideration of originality is more relied upon as to whether the object as a whole is sufficient to constitute a work. As a result, the ideas mentioned in the Copyright Law have gradually moved away from the original definition of the field of literature and art, and it is even easy for people to ignore an objective fact: ideas are the necessary conditions for works. The act of creating a work is actually a process of transforming ideas into expressions, and the "abstract filtering method" expounded by Judge Hand in the United States to distinguish between ideas and expressions is also based on the reconstruction of the creative process, that is, the process of continuously concretizing from the theme to the synopsis of the story, the design of the plot, the relationship between the characters, and the expression of words.

In fact, the structure of the causal relationship between thought and expression plays an important role in determining the originality of a work: "Thought and expression are interdependent, and 'expression' that does not reflect any thought is not the object of protection under copyright law." "In judicial practice, the parties and the court also take the causal relationship between thought and expression as an important basis to support the originality of the work. For example, in the copyright dispute case between Sankou Company and the Audiovisual Association, the court held that: "In order to fully express the ideological connotation of the song, the music video work involved in the case reflects the original labor of the producer through the tone and change of the characters' shapes, movements, sets, and shots, and through the combination of sound and painting with artistic beauty, and is a work created in a method similar to that of filmmaking." "The many ways of interpreting the connotation of originality, whether it is the author's personalized choice and judgment, or the labor and resources invested by the author, or whether it embodies the author's creative intention, are all attempts to demonstrate and explain whether the ideas are reflected in the work.The criterion for judging whether a new work based on the original work has originality is an attempt to establish a causal link between thought and expression. In human-led creative acts, the judgment of whether the corresponding expression reflects the idea is generally not triggered, and the focus of the argument is to distinguish the ideas and expressions in the elements of the work, and to provide a basis for the judgment of substantial similarity. However, this hidden causal structure between thought and expression still has its methodological value when judging whether the corresponding object should be protected by copyright.

4.3 Technological measures strengthen the exclusivity of copyright

Intellectual property has the characteristics of exclusivity, which is different from the exclusivity of tangible objects based on natural attributes, and the exclusivity of intellectual property rights is artificially created by the law. The exclusivity of copyright is mainly reflected in the exclusion effect of prohibiting others from using the work in a specific way according to the rights stipulated in the Copyright Law. On this basis, copyright does not bring direct and practical benefits to the right holder, but obtains the right to monopolize market transactions. Due to the immateriality and shareability of the work, the work is out of control after it is publicly published, and even if the unauthorized use of others falls into the scope of copyright, due to cost considerations, the copyright owner can only choose to tolerate some of the uses. At the same time, the exclusivity of copyright is also limited by public policies and transaction costs, and the legislation excludes certain acts from the boundaries of copyright rights through fair use on the grounds of promoting scientific research and large-scale use of digital works, and restricts the copyright owner's freedom to conclude contracts with statutory permission to achieve compulsory transactions. There are also theories that empower users of works and further limit the exclusivity of copyright.

The inherent defect of the inherent insufficiency of copyright exclusivity has been further amplified after the popularization and application of digital technology and network communication. Copyright owners have begun to apply technological measures to strengthen the exclusivity of their works, trying to control the unauthorized access, acquisition and use of works to the greatest extent. After the application of technical measures, circumvention tools and behaviors appear together, and it is difficult for the technical measures of the copyright owner to achieve the expected results. To this end, the Copyright Treaty and the Performances and Phonograms Treaty, concluded under the auspices of the World Intellectual Property Organization, impose obligations on members to protect technological measures in order to discourage circumvention of technological measures. After China included the relevant content of technological measures in the Regulations on the Protection of the Right of Information Network Transmission, the protection of technological measures was further added to the Copyright Law amended in 2020. As a result, two types of technical measures, "access control measures" and "copyright protection measures", are protected. The Copyright Law prohibits the act of circumventing or destroying technological measures, as well as the provision of equipment, components or technical services for the circumvention or destruction of technological measures.

"Copyright protection measures" are based on the specific rights expressly stipulated in the Copyright Law, supported by the exclusivity of copyright, and are a way to exercise rights. The purpose of the "access control measures" is to "ensure that copyright owners are able to benefit from the public's enjoyment of their works". The protection of technological measures in the Copyright Law provides incentives for copyright owners to strengthen their control over their works through technological measures, and thus the exclusivity of copyright has been substantially improved. The elevation of technological measures from a tool of private autonomy for individual use to a universal rule protected by law reflects the positive response of copyright law to technological change. From the perspective of the overall pattern of network governance, in the context of the improvement of platform review technical capabilities, the trend of strengthening platform responsibility is very significant. Through the design of privacy protection, the "gatekeeper clause" of personal information protection reflects the perspective of pre-engagement prevention and strengthens the exclusivity of vulnerable private rights. In the field of copyright, it is also necessary and realistic to require large network service providers with infringement risks and disputes to assume more active social responsibilities. The application of technical means can also play an auxiliary role in the internal institutional function of the correction and incentive effect of the Copyright Law.

5. Reconstruct the standardized path of AI-assisted generated content

In the context of the dominance of the cultural market by human-created content, the overall institutional structure of the Copyright Law realizes the incentive effect of maximizing the number of works, and the guiding function of the quality of works is realized through the market mechanism. But AI-assisted content has a significant advantage over the competition with human-created content. The incentive mechanism of the Copyright Law further strengthens this advantage, which also restricts the market adjustment mechanism of the quality guidance function of works. Although there is also an implicit normative value of corrective incentive effect within the Copyright Law, the judgment of the quality of works in the judicial link only has an impact on individual judgments, and the causal structure of thought and expression also plays a role in the post-event remedy, so it is difficult to form an overall ex-ante normative effect and cannot cope with the structural impact of AI-assisted generated content on the creative ecology. From the perspective of the overall trend of online platforms and AI governance, there is a certain consensus basis for strengthening the effective intervention of ex-ante regulation. It is necessary to appropriately move forward the implicit institutional function of the correction incentive effect of the Copyright Law, and make full use of technical means to form a systematic normative structure in the process of content generation, dissemination, and licensing.

5.1 The overall orientation of the Copyright Law in response to AI-assisted generated content

The application of artificial intelligence technology in the field of content generation has greatly improved the efficiency of learning and writing, and has a certain positive effect. However, the cost advantage and homogeneity of AI-assisted generated content will form a large-scale competitive advantage in the cultural market under the institutional effect of the general incentive of the Copyright Law, which is not conducive to the cultural ecology of encouraging high-quality creation. In the face of this conflict of interest and practical dilemma, it is necessary to rethink the legislative basis and value orientation of the Copyright Law, flexibly apply institutional tools, and form an overall idea to respond to AI-assisted generated content.

First of all, the primary legislative purpose of the Copyright Law is to encourage the creation of natural persons, and the causal relationship structure between ideas and expressions can directly exclude the content that does not have ideas from the scope of protection of the Copyright Law. If AI-assisted content is actually entirely made by AI or does not reflect human thoughts, it should not be incentivized by copyright law. However, the challenge in practice lies in how to effectively screen for it. AI-assisted content is difficult to identify from the appearance, especially many content directly generated by AI, which is very likely to claim copyright protection on the basis of AI-assisted generated content. Secondly, due to the limitation of the number of individual judgments, it is difficult to form an effective demonstration effect in the face of large-scale AI-assisted generation of content. The minimum amount of statutory damages may also incentivize infringement remedies for AI-assisted generated content, creating a judicial burden of large-scale mass litigation. Finally, AI-assisted content generation is mostly a secondary application of existing works, but its characteristics of intelligent plagiarism and the continuous generation and dissemination of homogeneous content make it difficult for the traditional infringement remedy model to play a practical role.

It must be admitted that AI-assisted generated content may also reflect high aesthetic value and cultural significance due to human intellectual investment, and the copyright law should not deny its copyrightability just because of the intervention of artificial intelligence. However, leaving the entrance to the copyright law also means that the regulation of AI-assisted generated content is in vain, and its cost advantage is in the competition of the cultural market, through the blessing of algorithm push and mobile Internet, it is easy to produce the negative impact of "bad money drives out good money". The experience of the copyright law in the protection of technological measures reflects the inherent rationality of technical solutions to technical problems, and it has become a realistic choice to identify AI-assisted generated content through technical means and improve the efficiency of judging others' intellectual investment in it. At the same time, benefiting from the huge traffic benefits brought by AI-assisted generated content, online platforms are also obliged to apply technical means to reduce the corresponding risk of copyright infringement and play their gatekeeper role in the communication process. In addition, as a means to ensure the normal order of the use of works and content transactions, it is also a supplementary way to correct the incentive effect of restricting the freedom of contracting of AI-assisted generated content that has reached a certain level of artificial intelligence, and evaluating the quality of its creation by way of statutory licensing.

5.2 Pilot mechanisms for intelligent identification and classification and labeling of AI-generated content on self-media platforms

After the application of generative AI, deep synthesis has gradually become the mainstream AI generation mode, which has greatly improved the efficiency and ability to integrate multimedia multi-platform data to form text, images, audio and video and other content, which has led to the risk of infringing on personal and property rights and interests, and may even endanger the country's political and economic order. The risk brought by deep synthesis mainly comes from its high simulation effect, which is difficult for the public to effectively identify and judge from the appearance level. Therefore, the corresponding governance path also focuses on how to establish an identification method that is convenient for tracking and timely intervention from a technical perspective, and form a content supervision mechanism involving multiple responsible entities. The European Union's Digital Services Act requires platforms to label deep synthesis, and China's regulatory legislation also reflects a similar line of thinking. The Provisions on the Administration of Algorithmic Recommendations for Internet Information Services, the Provisions on the Administration of Deep Synthesis of Internet Information Services, and the Interim Measures for the Administration of Generative Artificial Intelligence Services all set the obligation to label the content and information of deep synthesis for relevant service providers to a certain extent.

The obligation to label deep synthetic content is mainly to regulate the spread of false and harmful information, and to achieve the goal of safeguarding national security and protecting the public interest while protecting the legitimate interests of social entities. After editing, processing, integrating and reorganizing deep synthetic content, due to the involvement of human creative labor, it may be defined as AI-assisted generated content, and the regulatory method for deep synthetic content cannot be directly applied. Since it is difficult to effectively distinguish AI-assisted generated content from AI-generated content, the core issue involved in its subsequent dissemination and potential infringement disputes is how to select the creative contributions of others in it. To this end, on the one hand, we can learn from the labeling obligations set for deep synthesis content, and on the other hand, we should also explore more efficient identification methods in intelligent communication scenarios.

Currently, AI-generated content mainly involves text, images, audio, and video. As an important communication medium in the mobile Internet environment, the content and information released by the self-media platform are mainly presented in the aforementioned forms. As users of AI services, users of self-media platforms can directly upload AI-generated content, or disseminate it after certain editing and processing. In a copyright infringement dispute recently adjudicated by a Chinese court concerning an AI-generated image, the plaintiff claimed copyright protection for the image uploaded to the self-media platform, which was generated by the plaintiff using the corresponding software through text prompts. In the context of the country's increasingly perfect regulatory regulations on algorithm recommendation, deep synthesis and generative AI, we can try to realize mutual recognition of identifiers between self-media platforms and AI content generation software, and automatically analyze the proportion of AI-generated content through intelligent identification of uploaded content, and distinguish it into three categories: "AI-generated self-generated", "quasi-AI-generated self-generated" and "AI-assisted generated", and directly labeled. This distinction is only a preliminary presumption, and its purpose is to provide a typological analytical framework for subsequent dissemination of works, licensing transactions, and infringement remedies. AI-generated content cannot be protected by copyright, while AI-assisted generated content can be protected by copyright. In principle, quasi-AI-generated content cannot be protected by copyright, but its uploader can submit evidence to the platform, and if it can be proved that the creative contribution of the person in it has reached a corresponding proportion, it can apply for the labeling category to be adjusted to "AI-assisted generation". By exploring the mechanism of intelligent recognition and classification and annotation on the self-media platform, we will continuously strengthen the recognition ability of the system and the identification compatibility between various platforms, adjust and optimize the classification standards, and then try to further promote it to other fields.

5.3 Set copyright content filtering obligations for audiovisual content online platforms

As the most important institutional tool in the field of online infringement, the notice-takedown rule once contributed to the balance between the interests of the Internet industry and the copyright industry, but its shortcomings are becoming increasingly apparent. The cost of discovering infringements and issuing notifications on a continuous basis is getting higher and higher, and copyright owners are powerless to deal with mass infringements. Theoretically speaking, it is advocated to enhance the duty of care of network service providers, so that they can undertake the obligation of active review that matches the content management capabilities, and finally realize algorithmic copyright enforcement and improve the status quo of online infringement. Article 17 of the EU's Digital Single Market Copyright Directive substantially increases the copyright filtering obligations for internet service providers. At the level of judicial adjudication and administrative management, there has also begun to be a trend of requiring network service providers to undertake active review and filtering obligations.

The main purpose of setting copyright content filtering obligations for network service providers is to prevent large-scale infringement, and is committed to appropriately moving the link of infringement remedies forward through technical means, making up for the lag in rights remedies brought about by the private nature of copyright, and responding to the changes in the pattern of work dissemination in the mobile Internet environment. After a large amount of content can be generated by artificial intelligence, its cost advantage and the internal logic of algorithm push are more likely to stimulate the continuous generation and large-scale dissemination of homogeneous content, resulting in more serious infringement. The phenomenon of disseminating infringing content in the name of secondary creation is more common than directly disseminating infringing content without changing it. The infringer's assertion that its actions constituted fair use on the grounds of proper citation challenged the application of the notice-takedown rule. Due to the advantages of low cost and high efficiency of AI-generated secondary content, it is more necessary to actively filter and intercept it from the network platform level.

5.4 Establish a statutory licensing system for AI-assisted generated content

Through the intelligent identification and classification labeling mechanism, the content independently generated by artificial intelligence can be labeled, which is regarded as the content independently generated by artificial intelligence in the Copyright Law, and the corresponding reproduction and dissemination behaviors are not within the scope of adjustment of the Copyright Law. Pilot copyright content filtering mechanisms on audiovisual content online platforms, and through algorithmic copyright enforcement, the large-scale dissemination of homogeneous AI-assisted generated content can be avoided to a large extent. However, the lack of accuracy of intelligent recognition and algorithm filtering may lead to recognition errors or filtering errors, and there is also the possibility of readjusting quasi-AI-generated content to AI-assisted generated content, and after its public release, the public has a practical need to commercialize the relevant content.

For AI-assisted generated content that accounts for a relatively high proportion of the overall work, but has not yet reached the level of quasi-AI self-generated content, due to the low creative contribution of human beings, even if the unauthorized use constitutes infringement, it is difficult to obtain greater support in terms of damages when it finally enters the judicial evaluation process, and it is very likely that a large number of situations will apply the minimum amount of statutory damages. For example, in a case that supported the copyright protection of AI-generated images, the Chinese court only ordered the defendant to compensate the plaintiff for economic losses of 500 yuan. According to the traditional copyright trading and licensing model, users' doubts about the attributes of AI-assisted generated content works, as well as the need for large-scale continuous dissemination, will generate high communication and negotiation costs in terms of transaction conditions and transaction prices, and market failure will occur.

The statutory licensing system under the Copyright Law is mainly used to solve the problem of excessive transaction costs, which significantly increases the opportunities for the use and dissemination of works, and helps to achieve creative purposes. On the one hand, AI-assisted content has the possibility of large-scale and continuous dissemination, which increases the cost of negotiation and transaction, and on the other hand, even if it is determined to be infringing after entering the judicial process, the damages that can be obtained are limited due to the quality of creation, which increases the judicial burden and social costs. With this in mind, consideration could be given to introducing a statutory license for AI-assisted generated content. While increasing the opportunities for the use and dissemination of works, it also reduces the judicial burden, reduces the social cost of work licensing transactions, and appropriately moves forward the function of evaluating the quality of works. Specifically, on the basis of distinguishing between AI-generated content, quasi-AI-generated content and AI-assisted generated content, the proportion of AI-generated content in the overall work is determined. According to the type of work, the corresponding collective management organization shall determine the appropriate legal royalty calculation standard and calculation method, reflecting the principle that the higher the proportion of AI-generated content, the lower the royalty, and motivate natural persons to make greater creative contributions in the process of content generation.

conclusion

The primary purpose of copyright law is to encourage the creative acts of natural persons. Although AI-assisted generated content has the appearance of a work and there is a certain amount of human creative contribution in it, the role played by AI in the overall creative behavior cannot be directly equated with traditional creative tools. The institutional model of the Copyright Law to maximize the number of works will further strengthen the competitive advantage of AI-assisted generated content in the cultural market, lead to the large-scale dissemination of homogeneous creations, and make the copyright law's function of regulating and guiding the quality of works frustrated, endangering the overall creative ecology. Under the existing rules that AI-generated content cannot be protected by copyright, the response of the Copyright Law to AI-assisted generated content is, first, to seek an effective way to identify AI-generated content and determine the contribution of AI-generated content in the work, second, to prevent the massive application of AI technology from leading to the aggravation of large-scale copyright infringement in the online environment in the form of AI-assisted generated content, and third, to form appropriate prior guidance to encourage the creation of high-quality works, and to highlight the value of people's creative contributions in AI-assisted generated content。 Based on this, the relevant institutional suggestions are committed to facing the real impact of artificial intelligence technology on the creative field, reaffirming the primary value position of the Copyright Law to encourage natural persons to create, and pursuing the normative effect of maximizing the quality of works through the comprehensive application of technical means and public policies while maximizing the number of works.