The legal landscape for the use of AI in drug discovery in Indonesia, Thailand and Vietnam

Hien Thi Thu Vu
Tilleke & Gibbins, Vietnam
thuhien.v@tilleke.com

Adirek Siriwattasit
Tilleke & Gibbins, Thailand
adirek.s@tilleke.com

Wongrat Ratanaprayul
Tilleke & Gibbins, Indonesia
wongrat.r@tilleke.com

Rochmali Zultan
Tilleke & Gibbins, Indonesia
rochmali.z@tilleke.com

Artificial intelligence (AI) is reshaping the global pharmaceutical industry through its expanding role in drug discovery. From identifying novel compounds to enhancing clinical trial processes, its influence is increasingly evident in Southeast Asian countries, where the intellectual property (IP) regime faces new challenges posed by AI-generated pharmaceutical inventions. For legal practitioners, providing sound counsel in this evolving field requires a deep understanding of the IP framework and the uncertainties that surround it in Southeast Asia.

Indonesia

No dedicated AI law for drug discovery, yet

As of October 2025, Indonesia does not have a dedicated law specifically regulating the use of AI in drug discovery. The government’s current focus is on Fintech and digital transformation, with health and pharmaceuticals not yet the subject of AI-specific regulation. Instead, several general policy documents and guidelines touch on AI, such as:

• Ministry of Communication and Information (MOCI) Circular Letter No. 9 of 2023, which sets out ethical principles for AI use, including in regard to privacy, transparency and IP protection;
• the Financial Services Authority’s (Otoritas Jasa Keuangan or OJK) Ethical Guidelines on Responsible and Trustworthy AI in the Financial Technology, which apply to the use of AI in Fintech, not healthcare; and
• Press Council Regulation 1 of 2025, which covers the use of AI in journalism and only applies to media companies.

Draft national AI regulation

There is no dedicated law on AI in effect yet. Indonesia’s draft Presidential Regulation on AI is still under discussion. The draft Presidential Regulation is intended to address AI ethics, safety and liability, setting a basis for the development of future sector-specific regulations by the relevant ministries. Although originally expected to be implemented as early as the end of 2025, the government has pushed back the passing of the draft Presidential Regulation to early 2026.

Concurrently, a draft white paper on the national AI roadmap was published for public consultation in August 2025. This white paper is intended to help form the foundation of future national strategies and policies regarding AI governance in Indonesia.

IP: who owns AI-generated drug discoveries?

Indonesian law does not currently define or recognise AI-generated inventions within the IP framework. Patent law requires a ‘human inventor’, and AI is not considered as such. This means that:

• AI-generated molecules or drug candidates cannot be patented unless a human is involved in the inventive process; and
• copyright, trade secret and database protection may apply to AI-generated outputs (such as clinical trial data or algorithms), but only if the outputs meet the requirements of originality or secrecy, and the rights are held by the legal entity or person deploying the AI.

Legal uncertainties involving drug discovery

At present in Indonesia, the legal entity or individual generating AI data, models and outputs during drug research is considered to be the owner. Therefore, liability would likely fall on the legal entity or individual who deploys or controls the AI system. Prior art in Indonesia is defined as patent literature and non-patent literature disclosed in Indonesia or overseas. Based on this definition, AI-generated research outputs might be considered as able to defeat the novelty of a patent application for a new drug. Since there are no clear laws or guidelines, this remains an area of legal uncertainty.

Trade secrets: protecting AI-generated drug data

Trade secret protection does not require the involvement of a human as the author and, therefore, should be available for use in regard to AI algorithms, models and outputs (including clinical research results) if:

• the information is not generally known to the public;
• reasonable steps are being taken to maintain its secrecy; and
• the information has commercial value due to its secrecy.

This is a crucial consideration for pharmaceutical companies using proprietary AI models or data during drug discovery.

Practical guidance for drug discovery innovators

Until specific AI legislation is enacted, companies and researchers should:

• ensure human involvement in the inventive process to meet the current patent requirements;
• document the development process clearly, including the role of AI and human contributors;
• protect trade secrets and confidential data used in AI training and outputs through the creation of robust internal policies and contracts; and
• monitor related regulatory developments.

Looking ahead: ethics, collaboration and opportunities

Indonesia’s National AI Roadmap and ethical guidelines emphasise transparency, accountability, inclusivity and respect for IP rights. While these guidelines are not legally binding, they set the tone for best practice and the future regulation of drug discovery.

Indonesia is moving towards regulatory reform to address AI. By staying informed and engaged, stakeholders can help shape a legal framework that supports responsible AI innovation, while protecting the relevant rights and fostering trust in life sciences.

Thailand

The AI regulatory landscape

Despite AI’s growing prevalence, Thailand lacks standalone legislation dedicated to governing technology in general or specifically addressing AI-generated inventions, including drug discoveries. Several non-specific laws touch on AI, primarily the Computer Crime Act 2007 (B.E. 2550) and the Personal Data Protection Act 2019 (B.E. 2562). In addition, Thailand has introduced a planned statute circulated as Draft Principles of AI Law, which has been under consideration since May 2025.

The draft national AI law

Thailand’s draft AI legislation mainly focuses on promoting innovation and enforcing risk-based regulations to mitigate public harm, similar to the European Union’s AI Act. While not explicitly related to Thai patent or IP laws, the draft’s emphasis on safety, accuracy and human supervision may indirectly influence AI-driven research approaches in the context of drug discovery. The new legislation could lead to changes in future patent filing practices.

Ownership of AI-generated drug discoveries

Thai law does not recognise AI as having legal personhood, a core principle affirmed by the draft AI legislation. Although the Thai Patent Act 1979 (B.E. 2522) does not expressly define that the inventor must be a natural person, its language (eg, the use of ‘right’ in section 10 and ‘persons’ in section 15) implicitly indicates human agency and individuals capable of holding and transferring rights of ownership. This prevents any autonomous AI system from being designated as an inventor or holding ownership status for any drug discoveries.

Legal uncertainties involving drug discovery

Without AI-specific patent statutes and clear judicial guidance, questions about AI-generated drug discoveries are mainly answered using the Thai Patent Act. Some relevant examples are outlined below.

• Inventorship: Obtaining protection for AI-generated pharmaceutical solutions requires the human user and/or developer to be named as the inventor(s).
• Novelty and prior art: AI output could produce prior art that defeats the novelty of an invention. The gravity of this risk depends on the accessibility of the prior art, not its origin. According to section 6 of the Patent Act, an invention is not new if it was publicly disclosed or used before the filing date. AI output (eg, a novel pharmaceutical composition) becomes valid prior art when it is available in the public domain.
• Inventive step (obviousness): AI capabilities may have an effect on the Person Having Ordinary Skill in the Art (PHOSITA) standard in Thailand if it is presumed that the hypothetical PHOSITA is equipped with and is proficient in utilising AI tools. This enhanced PHOSITA could easily search the entire world for prior art and detect innovative synergistic effects, substantially raising the bar for obviousness. AI-generated inventions may then be considered obvious based on the higher standard and exposed to patent rejection decisions or invalidation. Thai courts have yet to rule on this possible enhancement, leaving the inventive step threshold for such inventions undefined.

Liability for AI infringements

The liability provisions set out in Thailand’s IP statutes consistently use the term ‘person’, which disregards autonomous AI systems. In the absence of relevant statutory rules on AI infringements, general tort principles may be invoked. This would assign responsibility to human actors (AI users, deployers or developers), holding them accountable for any unauthorised use or exploitation caused by the autonomous system by demonstrating their intent or implied intent concerning such use or exploitation.

Taking an alternative route: trade secrets

The Trade Secrets Act 2002 (B.E. 2545) offers a feasible alternative for protecting proprietary AI elements, such as algorithms, training datasets and clinical research results. Some relevant examples are outlined below.

• Requirements: A trade secret is defined as non-public information with commercial value that the owner has taken reasonable measures to keep confidential. Protection lasts so long as secrecy is strictly maintained, which requires pharmaceutical companies to implement robust confidentiality measures (eg, non-disclosure agreements and internal access controls).
• Clinical data: Clinical research results derived from an AI model can qualify as trade secrets if the underlying data and methodologies are confidential and commercially valuable.

Strategic recommendations

Considering the uncertainties and challenges, the legal strategy for AI-generated drug discovery in Thailand must rely on a dual-track approach. First, companies can patent the output (the novel drug) by meticulously documenting the human contribution to satisfy the inventorship requirement and monitor AI disclosures to prevent the creation of prior art that could defeat novelty. Second, companies must rigorously protect proprietary inputs, such as AI models, algorithms, training data and clinical research results under the Trade Secrets Act. This combined approach secures immediate, long-lasting protection for such assets as Thailand’s AI framework continues to evolve.

Vietnam

The regulatory landscape

Vietnam has laid the groundwork for AI governance through Decision 127/QD-TTg (2021) on the National AI Strategy. This Decision sets a vision for AI up until 2030, identifying AI as a key technology sector, establishing innovation centres and developing comprehensive legal frameworks.

Vietnam has yet to introduce a dedicated legal framework specifically addressing inventions generated by AI. In the absence of such targeted legislation, the country currently relies on a patchwork of general legal instruments, including the Civil Code 2015 and the Intellectual Property Law of 2005 (as amended in 2022) (the ‘IP Law’). Supplementary directives, such as Decision No. 6193/QD-SHTT (2021), Official Letter No. 1681/VPCP-KGVX (2024) and Decision No. 1290/QD-BKHCN (2024), offer guiding principles for AI technologies.

However, a transformative milestone was reached on 14 June 2025, when Vietnam’s National Assembly passed the Law on Digital Technology Industry. Set to take effect on 1 January 2026, this standalone law is exclusively focused on the digital technology sector, which includes regulatory provisions for AI systems. This bold legislative move signals Vietnam’s commitment to shaping a forward-looking legal landscape for emerging technologies.

Furthermore, Vietnam is drafting an AI Law, which is expected to be submitted to the National Assembly by the end of 2025 and take effect from 1 January 2026, if approved.

Additionally, an amended IP Law has been drafted, with the latest version of the draft submitted to the National Assembly on 27 October 2025. The amended IP Law is anticipated to become effective from 1 March 2026. Certain IP issues related to AI have been discussed during the drafting process of the amendment.

These regulatory developments show that Vietnam is taking important steps to build a clear and modern legal system for AI and IP.

AI-generated inventions and patent law

Inventorship

Under current Vietnamese law, only humans can be inventors. Article 122 of the IP Law defines an inventor as ‘the person who directly creates’ the industrial property subject matter. An invention is defined as ‘a technical solution in the form of a product or process aimed at solving a defined problem by applying the laws of nature.’ However, the Law does not specifically regulate whether such a technical solution may be created by AI.  An invention will be protected if it meets three conditions: novelty, inventive step and industrial applicability.

Article 6 of the IP Law further clarifies that industrial property rights are established on the basis of a protection title (eg, a patent) issued by competent authorities, with no provisions discussing AI inventorship.

During the drafting process of the amended IP Law, many versions of the draft reaffirmed that a patent will be refused or invalid if the listed inventor is not a human being. However, the final draft submitted to the National Assembly omits these clauses. Instead, a newly introduced ground for refusal or invalidation refers back to Article 122 of the current IP Law. This suggests an indirect interpretation that, if the inventor is not ‘a person who directly creates the invention, industrial design, or layout design’, the IP Office may reject the application or invalidate the granted patent.

It is also noted that the draft AI Law (which was published for public consultation in October 2025) stipulates that ‘the protection of intellectual property rights for inventions, technologies, works, and data related to artificial intelligence shall be carried out in accordance with intellectual property laws and other relevant legal provisions.’

Although the draft laws may still change before they receive official approval and are implemented, it is clear that, in Vietnam, AI is legally treated as a tool, not an autonomous inventor.

Patentability and prior art

According to Articles 60 and 61 of the IP Law, the patentability of an invention is established when comparing it with prior art that has been publicly disclosed in any form before the filing date or priority date. It is not stipulated in any legislation that prior art must be created by a person. Here, AI-generated outputs present challenges. The internet enables the online publication of a vast number of AI-generated documents, raising the prospect that AI-produced references could be treated as prior art, thereby defeating the novelty and inventive step of an invention.

However, public disclosure or being widely known has been the critical determinant for an AI-produced document in order to qualify as prior art. It is worth noting that, under the draft AI Law, one of the obligations imposed on AI systems is to clearly disclose (when content is made publicly available) whether it was generated by an AI system or significantly modified by one.

The presence of a large number of AI-produced prior art could impose additional burdens on patent offices during patent examinations and could lead to delays in granting patents.

Inventive step and the ai skills of a phosita

Article 61 of the IP Law determines the inventive step of an invention by comparing it with the knowledge of a PHOSITA in the relevant technical field to clarify whether an invention involves an ‘inventive step’.

The ability of AI technologies in the pharmaceutical sector to generate novel drugs raises questions about the continued relevance of the PHOSITA standard in evaluating the inventive step. Should the notional PHOSITA now be assumed to possess and skilfully employ AI tools? If the PHOSITA has access to AI, they might be able to combine known information in new ways easily, making some inventions seem obvious when they would not have been before, where previously inventors would have needed to conduct many experiments, tests or clinical trials, without big data, spanning multiple fields. This also raises the question of whether patent offices will use AI to examine the inventive step of an invention, whether it was created by humans alone or with help from AI.

This remains a complex issue. It is anticipated that, in the near future, the IP Law and any amendments will likely remain silent on the examination of the inventive step in the context of inventions involving contributions by AI.

Liability for autonomous AI infringements

Vietnamese law does not recognise AI systems as legal persons. Under Vietnamese civil law, specifically the Civil Code, a legal subject must be an individual or an organisation. Consequently, machines or computer programs, including AI, are not recognised as legal entities. This means that determining the legal status of AI as a subject under the law remains impossible. Thus, liability for patent infringements caused by AI defaults to human stakeholders, for example, developers and users.

Developers may be liable if their models are trained or deployed based on infringing datasets. Users may be responsible when exploiting AI outputs that encroach upon protected works. In certain cases, AI-generated products may be treated as ‘income’ under Article 109.2 of the Civil Code, so products generated by AI may be considered as ‘income’ brought about by the exploitation of AI.

Vietnamese law currently has no specific regulations on liability for damages related to AI. However, existing laws contain provisions that may address compensation for damages related to AI, such as the Law on Product Quality 2007, amended in 2018, and the provisions on non-contractual compensation set out in the Civil Code. When an AI-integrated product violates quality standards, the manufacturer is liable for compensation. If AI entities are considered assets (IP and tangible assets), according to Article 584.3 of the Civil Code, the owner or possessor is liable for damages. However, these provisions may only apply in certain cases or in regard to simple relationships, while more complex situations are harder to regulate. The draft AI regulations are silent on liability for autonomous AI infringements.

The protection of AI assets

Vietnam recognises trade secrets under the IP Law (Article 84), provided the information:

• is not publicly known;
• has commercial value; and
• is subject to reasonable confidentiality measures.

This framework is particularly relevant for AI models, algorithms, training data and clinical research results. However, unlike some jurisdictions, Vietnam has not yet established binding obligations requiring developers to ensure data legality, relying instead on voluntary guidelines promoting responsible AI development.

Practical guidance

Vietnam is positioning AI as a strategic technology, while drawing a firm line against recognising AI as an inventor. This dual approach, encouraging AI as a tool while preserving human authorship, creates a legal environment marked by both opportunities and uncertainty. For innovators in pharmaceutical research, the path forwards lies in combining traditional IP protection with proactive strategies, for example:

• ensuring human involvement in drug inventions to satisfy the inventorship requirements;
• documenting the role of AI and human contributors; and
• protecting sensitive inputs and outputs through robust trade secret management.

These strategies can assist drug developers in protecting their IP amid a changing legal landscape.