You are using an outdated browser and your browsing experience will not be optimal. Please update to the latest version of Microsoft Edge, Google Chrome or Mozilla Firefox. Install Microsoft Edge

October 26, 2025

The IP Puzzle of AI-Generated Songs: Protection, Responsibility, and the Future of Music Law

AI-generated songs are now making waves in Vietnam on platforms like TikTok, with tracks such as “Say mot doi vi em” quickly gaining popularity and sparking widespread attention. This phenomenon raises a host of legal and ethical questions: Who is the author of these songs? Can they be protected by copyright? Who is responsible if there is an infringement? These questions are becoming increasingly urgent as AI music becomes more mainstream in Vietnam.

Copyright Protection for AI-Generated Music in Vietnam

Under current Vietnamese law, copyright protection is reserved for works that bear the mark of human creativity. The 2022 amendments to Vietnam’s Intellectual Property Law reaffirm that only works created by humans are eligible for copyright. In practice, if a human meaningfully contributes to the creative process—by providing prompts, making selections, editing, or arranging—their contribution may be protected. However, if a song is generated entirely by AI without significant human input, it is unlikely to qualify for copyright protection.

When an AI-generated song does not qualify for copyright protection, the question arises as to whether the person who writes the prompts, edits, or compiles the work can still be considered the owner of an asset under the Vietnamese Civil Code. According to Article 105 of the Civil Code 2015, assets include objects, money, valuable papers, and property rights. While AI-generated music that is not protected by copyright is not considered money or valuable papers, it may be regarded as an object (in the form of a digital file or recording) or as a property right if it can be possessed, used, transferred, or exploited for value.

Use of AI-Generated Works Without Copyright Protection

If a song is not protected by copyright, does that mean anyone can use it freely? Not necessarily. The absence of copyright does not mean the work is entirely free of restrictions. Terms of service from AI platforms may limit commercial use, require attribution, or impose licensing fees. Other rights may also apply. The person who creates, edits, or compiles the AI-generated work may establish civil ownership over the digital file or recording as a type of digital asset, provided that the creation and use of the asset are lawful and do not infringe on others’ rights. This ownership is not the same as copyright, but it allows the owner to possess, use, and dispose of the asset within the limits of the law and any relevant agreements.

Additionally, laws against unfair competition, impersonation, or violations of personal rights (such as voice, name, or image) may still be relevant.

Voice Cloning and Related Risks

One particularly thorny issue is voice cloning. In Vietnam, performers’ rights protect both live performances and recorded voices. More importantly, copying or imitating a singer’s voice can primarily infringe upon the moral rights and personal rights of individuals as recognized under the Civil Code, which increasingly treats voice as a personal identifier. The main legal risk is the violation of moral rights, such as the right to protect the integrity and authenticity of one’s voice and the right to be recognized as the owner of that voice.

Best practices include obtaining written consent from the person whose voice is used, labeling content as “AI voice,” and avoiding any suggestion that the artist participated in or endorsed the work.

Similarity to Existing Works

Another significant risk arises when AI-generated music or lyrics resemble existing works. The standard for infringement is “substantial similarity” and access to the original. If an AI creates a segment that is sufficiently similar to a prior work, using that segment in a new recording or arrangement may constitute infringement. The fact that the AI was trained on large datasets does not exempt the output from scrutiny. Even if the input data was lawfully obtained, the output must still avoid copying protected material. Defenses such as coincidence, common style, or minor excerpts are assessed on a case-by-case basis, often requiring expert analysis.

Responsibility and benefit-sharing in the AI music ecosystem are complex. Users who prompt, select, edit, or publish AI-generated music are directly responsible for the outputs they release or exploit. AI platforms may also bear responsibility if they provide infringing tools or models, or fail to remove infringing content. Those who invest in, release, or commercially exploit AI-generated music may profit under contract, but they also assume corresponding legal risks, including compensation, takedown, or recall obligations.

Practical Recommendations for Creators and Publishers

For creators and publishers, several practical recommendations emerge. It is important to document the human role in the creative process, demonstrating selection and editing to support claims of authorship.

  • Similarity checks should be conducted using melody and lyric analysis tools, and expert opinions should be sought when necessary to avoid recognizable copying.
  • Voice governance is critical: Do not clone an artist’s voice without written consent, label AI-generated voices clearly, and avoid implying artist involvement.
  • Use models and training data with clear provenance, and ensure all samples, loops, and plugins are properly licensed, keeping records to prove origin.
  • Internal contracts should allocate rights and responsibilities among authors, producers, singers, engineers, and publishers, including indemnity clauses for intellectual property claims.
  • Platform terms should be reviewed carefully for output usage rights, commercial restrictions, and labeling obligations.
  • Finally, establish procedures for receiving and responding to takedown notices promptly to minimize damage.

Legal Outlook in Vietnam

Vietnam is actively shaping its legal framework to address the rapid growth of artificial intelligence. A draft Law on Artificial Intelligence released by the Ministry of Science and Technology is scheduled to take effect on January 1, 2026. In parallel, Vietnam’s Intellectual Property Law is under review, with discussions focused on how to accommodate AI-generated content.

The current framework does not recognize AI as an author, meaning that works created solely by AI may not qualify for copyright protection unless there is identifiable human contribution. This legal gap has prompted calls for clearer definitions of authorship, ownership, and liability in the context of AI-assisted creativity. While it is still too early to predict whether Vietnam’s upcoming laws will restrict or encourage AI-generated music, the direction appears cautiously optimistic. The government is balancing innovation incentives with ethical and legal safeguards, aiming to foster responsible AI development while protecting creators and consumers.

Key IP Takeaways on AI-Generated Content

AI-generated music challenges traditional IP frameworks on three fronts: authorship and protection, risks of voice cloning and similarity to prior works, and the allocation of liability among users, platforms, and publishers.

The safest path forward is proactive: Document the creative process, clear rights diligently, and use contractual safeguards. With these measures, businesses can treat AI not as a legal hazard, but as a sustainable creative tool in Vietnam’s fast-evolving music landscape.

As Vietnam’s evolving legal landscape offers both opportunities and uncertainties, stakeholders in the music and creative industries should stay informed and engaged with these regulatory developments to navigate the future of AI-generated content.

This article first appeared in Managing Intellectual Property.

RELATED INSIGHTS​ 

March 13, 2026
Vietnam’s Law on Intellectual Property (IP Law) has undergone continuous amendment in recent years, with the latest amendment issued at the end of 2025. Among the amended and supplemented provisions, the regulation that has perhaps attracted the most attention is a provision relating to the use of protected IP objects by artificial intelligence (AI) systems. Specifically, Article 7 of the 2025 IP Law introduces a completely new Clause 5, which reads in full as follows: “Organizations and individuals are permitted to use texts and data relating to intellectual property objects that have been lawfully published, and which the public is allowed to access, for the purposes of scientific research, experimentation, and training of artificial intelligence systems, provided that such use will not unreasonably affect the legitimate rights and interests of the authors and intellectual property rights holders in accordance with this Law. With respect to texts and data that are objects protected by copyright and related rights, the use of the texts and data as set forth herein must also be in accordance with the regulations of the Government.” Analyzing this newly added provision in the context of how it was conceived, as well as the challenges that still lie ahead, can provide some interesting insights. From Aspirations to Flight in Science and Technology From the end of 2024 and throughout 2025—the 50th anniversary of the country’s reunification—Vietnam witnessed numerous sweeping changes in many areas, including legislative development. It could be said that no sessions of the National Assembly have ever adopted as many laws, resolutions, and major policies as this one. The aspirations of the highest-level leadership have been concretized into major law and policy projects, which were drafted, developed, and passed at record speed. All of this was aimed at building a foundation for Vietnam to achieve
March 12, 2026
Thailand’s AI legislative framework took another step forward when the Office of the Consumer Protection Board (OCPB) issued a notification establishing guidelines for AI-generated advertising that may cause material misunderstanding about products or services. The notification, which is already in effect, was issued under the Consumer Protection Act B.E. 2522 (1979) and its amendments, which prohibit advertising that is unfair to consumers or may cause harm to society, including false or exaggerated statements and statements that may cause material misunderstanding about products or services. The notification addresses emerging advertising practices, including the use of images edited using software or AI to attract consumer interest or build credibility. The OCPB noted that such advertising may result in consumers misunderstanding the essential characteristics, condition, or usage of products, which violates consumer rights and causes damage. Key Requirements on AI-Generated or Digitally Manipulated Advertising Content For advertisements using still images or videos created or edited with software programs or AI tools that may cause the depicted product or service to differ from the actual product sold or service provided—which may cause misunderstanding regarding the condition, quality, quantity, or other essential aspects of the products or services—advertisers and business operators must comply with the following requirements: Prior authorization. Obtain approval from relevant regulatory authorities where required by law. Accurate representation. Ensure that the advertised size, quantity, volume, number, or composition matches the actual product or service being sold, whether in still images or videos. Mandatory AI disclosure labels. Display clear disclosures when AI or software is used to create or edit images, such as: “Real image or simulation edited using AI” “Photo from actual location or simulation edited using AI” “Photo from actual product or edited simulation” “Image created by AI” “Video created by AI” Clarity of disclosure. Ensure disclosures are clearly visible,
March 10, 2026
Thailand’s Ministry of Finance and Securities and Exchange Commission (SEC) have issued regulations broadening the criteria for determining who qualifies as a “major shareholder” of licensed securities and digital asset business operators. Under relevant SEC regulations, major shareholders of a regulated entity must obtain regulatory approval and undergo screening by the SEC. The revised framework introduces both shareholding-based and control-based tests to determine which shareholders require regulatory approval for a wider range of indirect ownership structures and de facto control. The Ministry of Finance notification took effect on February 21, 2026, while the SEC’s clarifying rules took effect on March 4, 2026. These changes aim to enhance transparency around beneficial ownership and strengthen regulatory oversight of entities operating in Thailand’s capital markets. Expanded Definition Under the revised framework, a “major shareholder” now includes persons who directly or indirectly hold more than 10% of the voting rights in a regulated company, as well as persons who exercise control over the regulated company or its shares. This system of two separate tests, based on both shareholding and control, differs from the prior regime, which focused primarily on shareholding thresholds and applied a more limited method for determining indirect shareholdings. The two tests (detailed below) operate independently of each other, and any person identified by either of the tests will be deemed a major shareholder. Shareholding-Based Test Broadens Indirect Ownership Attribution For the shareholding-based test, the SEC recognizes two existing methods for identifying indirect ownership, together with a new proportional attribution method. Any person captured under these methods, which are described below, will be regarded as a major shareholder of the regulated company and must obtain SEC approval as a major shareholder. First, the existing framework continues to apply to both first-tier and chain ownership structures. Approval is required for (1) first-tier
March 6, 2026
Thailand’s Legislation Consideration Committee of the Ministry of Interior has ruled that in-game loot boxes in online games do not constitute gambling under the Gambling Act B.E. 2478 (1935). This first-of-its-kind ruling provides useful guidance for online game operators and digital entertainment companies operating in Thailand. Background The ruling came in response to an inquiry concerning an online role-playing game operator that launched a campaign featuring a loot box mechanism. The mechanism allowed players to purchase a token in exchange for the opportunity to receive a virtual loot box containing randomized in-game items. The key features of this were as follows: The items received were digital, noncash items usable only within the game. The items could not be exchanged, redeemed, or converted into cash with the game operator. Items may differ in rarity but remain purely virtual. The central question was whether paying money to obtain randomized in-game items constituted a risk-based activity involving the chance to receive money or property of monetary value, which would constitute gambling under the Gambling Act. Committee Ruling The committee reached the following conclusions regarding the characteristics of the game’s loot-box mechanism: No cash or monetary equivalent: Players did not receive cash or property that could be exchanged for cash. The in-game items were merely usage rights within the online game ecosystem. No real-world monetary valuation: There was no determination of item value in real currency, and no mechanism for redeeming or converting items into money with the game operator. Any off-platform trading of in-game items between players is irrelevant to online game operators, as any value arising from such transactions is determined by the market rather than by the operators themselves. Service fee characterization: Payments made by players purchasing in-game loot boxes constituted fees for online game services. Accordingly, the committee concluded