Applicants seeking patent protection for AI-related inventions in the Philippines must ensure that the application clearly identifies the human inventor and sufficiently explains the technical contribution of the AI-related invention. Rowen D. Palma and Ninna Joyce P. Delantar explain.

Artificial intelligence is becoming increasingly important in the development of new technologies and is now used both to develop AI technologies and to provide technological solutions in fields such as healthcare, manufacturing, finance and transportation. Its growing use in research and development has raised new questions for patent systems, particularly where AI is used to create or assist in developing an invention or forms part of the invention itself. These issues are particularly relevant to applicants seeking patent protection for AI-related technologies across multiple jurisdictions.

The Philippines has likewise begun to address these issues. The Intellectual Property Office of the Philippines (IPOPHL) has recently published the Examination Guidelines of Inventions, Utility Model, and Industrial Design Applications Relating to Artificial Intelligence, dated July 2025 (the AI examination guidelines). The guidelines supplement the existing Philippine patent examination guidelines, including the Revised Manual for Patent Examination Procedure and the Patent Examination Guidelines for Information Communications Technology and Computer Implemented Inventions (the ICT and CII guidelines).

For applicants considering patent protection for AI-related inventions in the Philippines, the guidelines provide useful guidance on how the IPOPHL will approach issues such as inventorship, patentability, disclosure and enablement and claim clarity.

AI inventorship: The DABUS cases and the Philippine position

One of the key questions in AI and patent law is whether an AI system can be recognized as an inventor. This issue has been, and in some jurisdictions continues to be, considered in the DABUS series of cases. These cases involve DABUS, an AI system developed by Dr. Stephen Thaler, which was identified as the sole inventor of two inventions: a fractal-based food container and a flashing light beacon for emergency use. Since then, courts and patent offices in several jurisdictions have considered whether AI can be recognized as an inventor.

The position adopted in the major DABUS cases has generally been that, under existing patent laws, an AI system cannot be an inventor. In the United States, the U.S. Court of Appeals for the Federal Circuit in the case of Thaler v. Vidal upheld the U.S. Patent and Trademark Office’s refusal of both DABUS applications on the ground that an inventor as defined under the U.S. Patent Act refers to “the individual who invented or discovered the subject matter” and that the term “individual” refers to a natural person. As such, under U.S. law, ‘‘only a natural person can be an inventor, so AI cannot be.’’ The U.S. Supreme Court subsequently declined to review the case when it was appealed on certiorari.

Australia provides a notable development in the DABUS cases. Its Federal Court initially held that AI could be recognized as an inventor because the Australian Patents Act did not expressly require an inventor to be a natural person. However, this decision was eventually reversed by the Full Federal Court of Australia, which held that due to the statutory language, history and purpose of the Australian Patents Act, the term “inventor” could only refer to a natural person.

The European Patent Office (EPO) has likewise maintained that a patent application must identify a human inventor. The original DABUS applications were refused because DABUS was identified as an inventor instead of a natural person as contemplated under the European Patent Convention. In a later divisional application, Thaler identified himself as the inventor but submitted an addendum which stated that DABUS had autonomously conceived the invention. The EPO refused the divisional application which was upheld by the EPO Technical Board of Appeal, emphasizing that the inventor must be clearly identified as a human being.

Japan has also addressed the issue. In January 2025, the Japan Intellectual Property High Court upheld the rejection of a DABUS application, concluding that the current Japanese patent framework recognizes natural persons as inventors. The same court also noted the need for legislative consideration of the implications of AI-generated inventions.

In view of these international developments, the IPOPHL’s AI Examination Guidelines take a similar position. The guidelines categorically provide that AI cannot be named as an inventor, a maker, or a designer in the Philippines. While the Philippine Intellectual Property Code does not specify that an inventor must be a “natural person”, AI nevertheless cannot be recognized as a “person” capable of becoming an inventor for the following reasons:

1. Under Philippine law, a “person” may refer to either a natural person, which is a human being, or a juridical person which acquires personality as provided under law.
2. Not being a “person”, AI does not have the civil personality to exercise acts of ownership and dominion over the patent, which is the main purpose of patent ownership and registration under the IP Code.
3. AI cannot comply with the requirement of identification, particularly by providing the inventor’s legal name, in an application, as required by the Philippine implementing rules.

The implication is straightforward in the Philippines. An AI system should not be identified as the inventor in a Philippine patent application. Where multiple inventors are identified, the guidelines likewise indicate that joint inventors, makers and designers must all be human beings.

AI-assisted inventions: AI as a tool

Nonetheless, the guidelines expressly recognize AI as a tool that may assist inventors, makers and designers. Where AI is used to generate an invention or design, the person who provides the prompts or instructions to the AI would still be considered the inventor and/or the applicant. This is particularly relevant to businesses that use AI tools as part of their research and development processes.

The guidelines, however, caution that extensive use of AI may give rise to other legal considerations. Although the IP Code does not limit the extent to which AI may be used in generating an invention or design, such use may have legal consequences under other applicable laws, such as copyright, unfair competition and infringement.

IPOPHL examiners are also advised to exercise additional diligence when examining AI-assisted inventions. As AI systems may generate outputs based on unreliable or fabricated sources, examiners are instructed to raise such findings in an office action to maintain the quality and integrity of patents granted by the IPOPHL. Applicants are therefore advised to take care to verify the sources and technical information referred to by their AI-assisted inventions or designs.

Subject matter patentability of AI-related inventions

While computer programs per se are statutorily excluded from Philippine patent protection, the Philippine implementing rules and the ICT and CII guidelinesrecognize that computer-implemented inventions may still be patentable if they provide a “technical contribution” or “technical effect” over the known art. The invention must therefore produce a concrete and external technical effect beyond the mere execution of a program or mathematical calculation on a computer system. This principle likewise extends to AI-related inventions, which the AI examination guidelines treat as a sub-category of computer-implemented inventions.

In determining the subject matter eligibility of an AI-related invention, the AI examination guidelines lay down the Five-Step Claim Analysis, which was first introduced in the ICT and CII guidelines, as follows:

1. Does the claim involve computer-implemented inventions and AI?
2. Is the claim directed to matter which is against public order or morality?
3. Is the claim directed to a statutory class of invention i.e. product or process within the meaning of the IP Code?
4. Does the claim involve non-technical matters?
5. Does the claim constitute technical character after weighing all the factors?

The step which invites a higher degree of scrutiny particular to AI-related inventions lies in Step 5, namely, in the determination of “technical character”.

For product claims, the presence of man-made tangible embodiments with a real-world use or the application of computer-related aspects and tangible features to realize the invention and produce technical effect can be evidence of technical character.

Process claims, on the other hand, must involve a recitation of the machine or the use of a machine to implement the claimed steps. The claims must implement a concept in some tangible way and must not be a mere statement of a concept. In practical terms, the claimed AI-related features should therefore be implemented through, or in conjunction with, appropriate hardware components rather than being claimed solely as an abstract AI concept or computer program.

Disclosure and enablement requirements

The guidelines classify AI-related inventions into (i) core AI technologies, or inventions which advance AI technology itself, and (ii) applied AI technologies, which apply existing AI technologies to solve technical problems in a particular field.

As core AI technologies seek to build on the foundation of AI technology itself, the guidelines impose a higher level of disclosure for such inventions. An application involving core AI technologies should disclose a detailed description of the AI system, a description of the training data employed, the relationship between the input data and output and the hardware and software implementation details. The disclosure must be reproducible by a person skilled in the art without undue burden or inventive effort nor must it amount to an invitation for further research.

On the other hand, applied AI inventions must show how AI is integrated to solve technical problems in specific fields. An application for an applied AI invention should identify the specific technical field and problem addressed, explain the AI’s integration with other components of a system, describe the relevant input data and the role of the AI model and demonstrate the resulting technical effect through, where appropriate, benchmarks, simulations or real-world test results. Likewise, the disclosure must also be sufficient to allow the invention to be reproduced using known or reasonably available tools and models.

For both types of AI-related inventions, the guidelines require disclosure of the AI system’s training data sets, training model or algorithm, and its technical effect to ensure its reproducibility. While the training dataset itself need not be disclosed, the characteristics of the training data which are required to produce the technical effect must be described in sufficient detail, including its quality, quantity, scope and method by which the data was obtained. As to the AI system’s training model, the computational model specifications, algorithms and specific implementations of the model must be described for a person skilled in the art to evaluate its technical features. Finally, the technical effect of the AI invention and how it is achieved with the claimed AI must be clearly disclosed through the provision of experimental data, mathematical proofs or other relevant evidence.

The guidelines also stress the difference between “prophetic examples” or hypotheticals from actual working examples generated by AI tools, with only the latter accepted as proof of enablement in a patent application.

Clarity of claims in AI-related inventions

The guidelines stress that for AI-related inventions, clarity is essential due to the technical complexity and potentially abstract (i.e. “black box”) nature of AI systems and methods. The general requirements for clarity of claims under existing Philippine patent rules continue to apply to AI-related inventions, while the guidelines provide additional considerations specific to AI. Thus, AI-related claims must be technically precise, definite in scope, functionally defined, consistent in language and free from ambiguity. Examiners are advised to pay particular attention to undefined AI-related buzzwords such as “AI module,” “intelligent data processing” or “smart decision-making”. Functional claim language is still required to be supported by sufficient technical disclosure to explain how a claimed feature is achieved, rather than merely stating a desired result.

Practical takeaways for applicants

For applicants seeking patent protection for AI-related inventions in the Philippines, the key consideration is to ensure that the application clearly identifies the human inventor and sufficiently explains the technical contribution of the AI-related invention. Applicants should ensure that the specification of their application provides enough technical detail which, in summary, would enable the reproduction of the AI-related invention. Finally, the claims should clearly define the technical features and implementation of the invention, rather than relying on broad or purely functional AI terminology. For more guidance, the guidelinesinclude examples of common issues and illustrative cases showing patentable claims for AI-related inventions.