AI and Copyright in Thailand: Is AI-Generated Work Protected?

Last updated 8 November 2026 · Legal status: the Copyright Act B.E. 2537 (1994) is in force (consolidated text as amended, effective from 23 August 2022 (B.E. 2565))¹⁹ · Thai copyright law contains no provision on artificial intelligence

Download this article as a PDF

I. AI and Copyright in Thailand: Is AI-Generated Work Protected?

Under Thai law, a work generated by artificial intelligence on its own, without human creative input, is unlikely to be a copyright work, because Section 4 of the Copyright Act B.E. 2537 (1994) defines an “author” as the person who makes or causes to be made a creative work, and Section 8 grants copyright to the author under conditions that apply to persons. Where a human uses AI as a tool and puts their own creativity into the work, the human creative contribution remains protected.

The question arises in businesses of every size. Marketing teams use image generators for advertising, content producers use text generators to draft articles, and software developers use coding assistants. When others copy that material, the business needs to know what rights it actually holds. When the business wants to train a system on other people’s works, it needs to know whether that use infringes anyone’s rights.

The Copyright Act B.E. 2537 (1994), as amended by the Copyright Act (No. 2) to (No. 5), most recently No. 5 of B.E. 2565 (2022), contains no provision that refers to artificial intelligence. The answer therefore has to be built from the general principles of the Act: the definition of “author”, the nature of a “copyright work”, the conditions for acquiring copyright, and the term of protection. This article explains those four principles from the statutory text, then compares the approach of the United States, the United Kingdom and the European Union, where courts and legislators have answered the question more directly.

II. Thai Law: Author, Copyright Work and Copyright Owner

The Copyright Act B.E. 2537 (1994) vests copyright in the “author”, whom Section 4 defines as the person who makes or causes to be made a creative work. A protected work under Section 6 must be a creative work of a type listed in the Act, and Section 6 paragraph two excludes ideas, procedures and principles from protection, so that only expression is protected.

1. The author under Section 4

Section 4 defines an “author” as a person who makes or causes to be made any creative work that is a copyright work under the Act, and defines “copyright” as the exclusive right to do any act under the Act in relation to a work the author has made.¹ Both definitions tie copyright to the person who makes the work. The Act does not grant copyright to a work in the abstract, without an author.

As a matter of interpretation, the words “causes to be made” are wider than making the work by hand. A person who creates a work with a device or tool, such as a photographer with a camera or a designer with graphics software, still causes the work to be made, because the choices of composition, viewpoint and arrangement are that person’s decisions. The question for AI is whether the user decides enough about the expression of the work to be the person who causes it to be made. This has to be assessed work by work.

2. Copyright works under Section 6 and what is not protected

Section 6 paragraph one provides that copyright works are creative works in the categories of literary works, dramatic works, artistic works, musical works, audiovisual works, cinematographic works, sound recordings, broadcast works, or any other work in the literary, scientific or artistic domain “of the author”. Paragraph two provides that copyright protection does not extend to ideas, procedures, processes or systems, methods of use or operation, concepts, principles, discoveries, or scientific or mathematical theories.²

Section 6 paragraph two bears directly on AI. Most text instructions typed into a system (prompts) describe the idea or approach of the work, such as “a seascape at dusk in the style of an oil painting”. Such an idea is not protected under paragraph two in any case. If the system determines the lines, colors, composition and every detail of the image, what the user contributed is an idea, while the expression is the output of the system. This is the same reasoning used by the U.S. Copyright Office, discussed in Part V.

3. Conditions for copyright under Section 8 and the term under Section 19

Section 8 makes the author the owner of copyright subject to conditions. For an unpublished work, the author must be a Thai national or reside in Thailand, or be a national of or reside in a country party to a copyright convention to which Thailand is also a party, throughout or for most of the period of creation. Where the author must be a Thai national and the author is a juristic person, that juristic person must be established under Thai law.³

Section 19 sets the general rule, subject to Sections 21 and 22, which set separate terms for certain types of work: copyright subsists for the life of the author and for fifty years after the author’s death. Where the author is a juristic person, copyright lasts fifty years from creation, or from first publication if the work is published within that period.⁴

Nationality, residence, death and incorporation are all attributes of natural persons or juristic persons. An AI system has no nationality, no residence, and no death from which to measure the term of copyright. In the writer’s view, the structure of Sections 8 and 19 shows that Thai law treats the author as a person. This view is an interpretation of the statutory text, and a Thai court will still have to decide the point in an actual case.

4. Who owns copyright when an employee or contractor creates the work

Section 9 provides that, for a work created by an author as an employee, unless otherwise agreed in writing, copyright belongs to the author, but the employer may communicate the work to the public for the purpose of the employment.⁵ Section 10 provides that, for a work created under a commission from another person, the person who commissioned it owns the copyright, unless the author and the commissioning party have agreed otherwise.⁶

These two sections matter when a business has employees or contractors use AI. Copyright can pass to the employer or the commissioning party only for the part that is a copyright work. If an image or a text was produced entirely by the system without any creative contribution from the employee, a contract assigning copyright does not create rights in that part. Employment and service contracts should therefore require the contractor to disclose which parts of the work used AI and which parts are the contractor’s own creation.

III. AI-Assisted Works: Which Parts Belong to a Human Author?

A work made with AI as a tool can still be protected in the parts created by a human: what the human wrote or drew personally, the creative selection and arrangement of outputs, and significant modification of outputs. Material that the system produced on its own from a text instruction (prompt) alone is unlikely to be the work of an author.

Separating the human contribution from the system’s contribution allows the copyright question to be answered in line with Sections 4 and 6. Common ways of using AI fall into three patterns.

• Pattern one: the system generates the whole work from a text instruction. The user types a description and uses the output without change. What the user contributed is an idea, which Section 6 paragraph two does not protect, and the expression comes from the system. A work of this kind carries a high risk of having no author in law.
• Pattern two: a human selects and arranges outputs. The user generates many images and selects and arranges them into a picture book, or compiles several text outputs into a new piece of writing. Selection and arrangement are human decisions. If they are sufficiently creative, the person who selects and arranges may be the person who causes the work to be made under Section 4, as to that selection and arrangement.
• Pattern three: a human creates the work and uses AI for some steps. For example, a writer writes the manuscript and uses a system to check grammar, or a designer draws a sketch and uses a system to color part of it before editing further. This is close to using an ordinary tool, and the user remains the author of what the user made.

As a matter of evidence, a person who claims copyright must be able to show which parts they created. A business that uses AI to produce material should keep a record of the work, such as each draft, the prompts used, the outputs received and the edits made afterwards. These records separate the human contribution from the system’s contribution if a dispute arises.

IV. Does Training AI on Others’ Works Infringe Copyright?

Thai law has no specific exception for training AI systems. Copying a copyright work to train a system may be reproduction under the definition in Section 4, which Section 27 treats as infringement if done without permission, unless it falls within an exception under Section 32, which requires that the use not conflict with the copyright owner’s normal exploitation of the work and not unreasonably prejudice the owner’s legitimate rights.

Section 15 gives the copyright owner the exclusive right to reproduce or adapt the work, to communicate it to the public, and to license others to exercise those rights.⁷ Section 4 defines “reproduction” to include copying by any means, making copies, or recording, from the original or from a copy, of a substantial part, whether in whole or in part.⁸ Section 27 provides that reproducing or adapting a work, or communicating it to the public, without permission under Section 15 (5) is an infringement of copyright.⁹

Training an AI system generally requires collecting large quantities of text or images and storing them in a computer system. That step involves copying whole works, so it may fall within the definition of reproduction in Section 4. The next question is whether the reproduction falls within an exception. Two provisions are relevant.

Section 32 paragraph one sets the principle that an act done to another person’s copyright work is not an infringement if it does not conflict with the copyright owner’s normal exploitation of the work and does not unreasonably prejudice the owner’s legitimate rights, and paragraph two, item (1), gives as an example research or study of the work “that is not for profit”.¹⁰ Training a system to provide a commercial service does not meet the condition “not for profit” in item (1), so it has to be assessed under the general principle in paragraph one, that is, whether the training conflicts with the copyright owner’s normal exploitation of the work. That depends on the facts of each case, for example whether the system’s outputs compete with the original works in the same market.

Section 32/2 exempts reproduction that is necessary in a computer system so that equipment, or the transmission of a work through a computer system, can function normally, provided the work was made or obtained lawfully.¹¹ This exception is aimed at technical copies made during use or transmission. Keeping a dataset for training a system is a different kind of use. In the writer’s view, Section 32/2 therefore does not cover collecting works to train a system.

Since Thai law has no specific rule, a business that trains or fine-tunes an AI system on others’ data should use licensed data, works whose copyright has expired, or its own data. It should also note that training data may contain personal data, which is subject to the Personal Data Protection Act B.E. 2562 (2019), separately from copyright. The lawful bases for processing personal data are explained on the page Lawful Bases for Processing Personal Data.

V. Comparative Law: the United States, the United Kingdom and the European Union

The United States requires a human author, as held in Thaler v. Perlmutter. The United Kingdom has a specific provision in Section 9(3) of the Copyright, Designs and Patents Act 1988. The European Union has a text and data mining exception that rightholders can opt out of, and it imposes copyright duties on providers of general-purpose AI models.

1. United States: Thaler v. Perlmutter and the Copyright Office report

The case began when the applicant sought to register copyright in an image that named an AI system called the Creativity Machine as its author. The U.S. Copyright Office refused registration, the applicant sued, and on 18 March 2025 the U.S. Court of Appeals for the District of Columbia Circuit affirmed the district court judgment upholding that refusal. The court held that, as a matter of statutory law, the Copyright Act requires all work to be authored in the first instance by a human being. In the same opinion the court stated that the human authorship requirement does not prohibit copyrighting work made by or with the assistance of AI.¹² The applicant petitioned the U.S. Supreme Court, which denied the petition on 2 March 2026.¹³

The U.S. Copyright Office published Part 2 of its report Copyright and Artificial Intelligence, on copyrightability, on 29 January 2025. The report concludes that, with generally available technology at that time, prompts alone do not provide sufficient control over the output; that human authors are entitled to copyright in their works of authorship that are perceptible in AI outputs, in the creative selection, coordination or arrangement of material in the outputs, and in creative modifications of the outputs; and that the use of AI tools to assist rather than stand in for human creativity does not affect the availability of copyright protection.¹⁴ This approach is close to the three patterns in Part III, and it rests on the same principle as Section 6 paragraph two of the Thai Act: the law protects expression, not ideas.

2. United Kingdom: computer-generated works under Section 9(3)

Section 9(3) of the Copyright, Designs and Patents Act 1988 provides that, for a literary, dramatic, musical or artistic work which is computer-generated, the author is taken to be the person by whom the arrangements necessary for the creation of the work are undertaken. Section 178 defines “computer-generated” as meaning that the work is generated by computer in circumstances such that there is no human author of the work.¹⁵

This provision differs from the U.S. approach. English law has accepted since 1988 that a work may have no human author, and it allocates copyright to the person who made the arrangements. Thailand has no comparable provision, so a Thai court cannot apply Section 9(3) directly. The provision does show that granting copyright in computer-generated works requires a specific statutory basis.

3. European Union: the text and data mining exception and the AI Act

Article 2(2) of Directive (EU) 2019/790 on copyright in the Digital Single Market defines text and data mining as any automated analytical technique aimed at analysing text and data in digital form in order to generate information such as patterns, trends and correlations. Article 4(1) requires Member States to provide an exception for reproductions and extractions of lawfully accessible works for the purposes of text and data mining, and Article 4(3) makes that exception apply only if rightholders have not expressly reserved the use in an appropriate manner, such as machine-readable means for content made publicly available online.¹⁶

Article 53(1)(c) of Regulation (EU) 2024/1689 on artificial intelligence (the AI Act) requires providers of general-purpose AI models to put in place a policy to comply with Union copyright law, in particular to identify and comply with reservations of rights under Article 4(3) of Directive (EU) 2019/790, and Article 53(1)(d) requires them to draw up and make publicly available a sufficiently detailed summary of the content used for training, according to a template provided by the AI Office. The obligations in Chapter V, which includes Article 53, apply from 2 August 2025 under Article 113(b).¹⁷

Thai businesses that supply data or AI services in the EU market should therefore check whether the providers of the models they use comply with these duties. Thai content owners who do not want their works used for text and data mining in the European Union may reserve their rights by machine-readable means under Article 4(3).

VI. Law and Economics and Interdisciplinary Perspectives

In law and economics, copyright balances the incentive to create works against the cost of limiting access to them. William M. Landes and Richard A. Posner set out this framework in their 1989 article An Economic Analysis of Copyright Law. The same framework helps explain why the laws of several countries still do not grant copyright in outputs that a system generates without human creative contribution.

Landes and Posner explain that creative works are expensive to create in the first place but cheap to copy. Without copyright, others could copy and sell a work without bearing the cost of creating it, and the incentive to create would fall. On the other hand, protection that is too broad raises costs for later creators who need earlier works as raw material. Good copyright law must find the balance between these two sides.¹⁸

Applied to AI: if a user can produce each output at a cost far lower than creating a work by hand, the argument for an exclusive right to recover the cost of creation carries less weight for outputs the system generates on its own, while granting exclusive rights in a large number of outputs may raise costs for other creators. Investment in developing the system has other channels of protection, such as copyright in computer programs, which Section 4 treats as literary works, and contractual arrangements. This analysis uses an economic framework to explain the direction of the law. It is not a conclusion that Landes and Posner drew about AI.

On training, the same framework explains the difference between Thailand and the European Union. A text and data mining exception lowers the cost of access to data for system developers, while the reservation right under Article 4(3) of Directive (EU) 2019/790 gives bargaining power back to rightholders who want to be paid. Thailand still relies on the general principle in Section 32, which leaves the court to weigh the matter case by case.

VII. Practical Points for Businesses That Use AI to Create Material

A business that uses AI to create material for commercial use should separate the human creative contribution from the system’s output, keep records of the work, address copyright and AI use in employment and service contracts under Sections 9 and 10, and check the AI provider’s terms before commercial use.

1. Identify the human contribution. Material the business wants to protect, such as a logo, key advertising copy or core brand images, should contain an identifiable human creative contribution, because material the system produced on its own may have no copyright that can be enforced against copiers.
2. Keep records of the work. Drafts, prompts, system outputs and edits are evidence of which parts are a person’s creation.
3. Write contracts that cover the point. Contracts with employees should address copyright in writing under Section 9. Contracts with contractors should address copyright under Section 10, with a duty to disclose AI use and a warranty that the work does not infringe the rights of others.
4. Check data used to train or fine-tune systems. Use licensed data or the business’s own data, and check personal data issues under the Personal Data Protection Act B.E. 2562 (2019).
5. Check the law of the target market. A business that provides services in the European Union or sells works in the United States must consider the rules of that jurisdiction, which differ from Thai law as described in Part V.

Putting contracts and records in place before work begins reduces what has to be proved when someone copies the work. The idea of using law to set up structures in advance is explained in the article Law as a tool for shaping the future (Thai), and the principles for interpreting contract terms, including terms on copyright, are on the page Good faith and contract interpretation (Thai).

VIII. Frequently Asked Questions

Frequently asked questions on AI and copyright include whether AI-generated work is protected, whether writing a prompt makes the user an author, who owns copyright when AI assists in creating a work, whether training AI on others’ works is lawful, and how long copyright lasts. Each answer cites Sections 4, 6, 9, 10, 19, 27 and 32 of the Copyright Act B.E. 2537 (1994) as relevant.

Is AI-generated work protected by copyright under Thai law?

The Copyright Act B.E. 2537 (1994) has no provision on artificial intelligence. Copyright vests in the author, whom Section 4 defines as the person who makes or causes to be made a creative work, and Sections 8 and 19 set conditions based on nationality, residence, death and juristic persons, which are attributes of persons. In the writer’s view, a work generated entirely by a system without human creative contribution is unlikely to be a copyright work, while the human creative contribution remains protected.

Does writing a prompt make the user an author?

Generally not by itself. Most prompts describe the idea of a work, which Section 6 paragraph two does not protect. If the system determines all of the expression, the user is unlikely to be the person who causes the work to be made under Section 4. The U.S. Copyright Office concluded in its January 2025 report that prompts alone do not provide sufficient control over the output. As a matter of interpretation, a prompt written with creativity may itself be a literary work, separate from the system’s output.

Who owns copyright when AI assists and a human edits the output?

A person who creatively edits, selects or arranges the output is the author of what that person contributed. If the person acts as an employee, copyright belongs to the author under Section 9 unless otherwise agreed in writing, and the employer may communicate the work to the public for the purpose of the employment. If the work is commissioned, the commissioning party owns copyright under Section 10 unless otherwise agreed. Material the system generated on its own remains outside protection.

Does training AI on others’ works infringe copyright in Thailand?

It may. Copying works to train a system may be reproduction under the definition in Section 4, which Section 27 treats as infringement without permission. Thai law has no specific text and data mining exception. Commercial training does not fall within the exception for research that is not for profit in Section 32 paragraph two, item (1), so it must be assessed case by case under the general principle in Section 32 paragraph one. Using licensed data or the business’s own data carries less risk.

Do other countries grant copyright to AI-generated works?

The United States does not grant copyright to works without a human author, following the D.C. Circuit decision in Thaler v. Perlmutter of 18 March 2025, which the U.S. Supreme Court declined to review on 2 March 2026. The United Kingdom has a specific rule in Section 9(3) of the Copyright, Designs and Patents Act 1988, which treats the person who made the arrangements necessary for creating a computer-generated work as its author. Thailand has no provision like the UK rule.

How long does copyright last in AI-assisted work?

Section 19 sets the general rule that copyright lasts for the life of the author and fifty years after the author’s death. Where the author is a juristic person, copyright lasts fifty years from creation, or from first publication if the work is published within that period. The rule is subject to Sections 21 and 22, which set separate terms for certain types of work. The term applies to the part that is the author’s work. Material a system generated without an author has no copyright to measure.

IX. Related Topics across the Eksiam Group

Copyright in AI-assisted work is connected with contract law, company law and personal data law, in particular contracts that must address copyright under Sections 9 and 10 of the Copyright Act B.E. 2537 (1994), and the use of personal data as training data. The following pages build directly on this article.

• Lawful Bases for Processing Personal Data — when data used to train or prompt an AI system includes personal data
• Personal data: property or a right? (Thai) — the legal status of data used as raw material for systems
• Good faith and contract interpretation (Thai) — interpreting contract terms on copyright and AI use
• Eksiam Corporate Law — drafting employment and service contracts that address copyright
• Eksiam Legal practice areas