By using an increasing amount of generative AI to produce advertising copy, product designs and other content, businesses are opening themselves to a variety of legal challenges. Ma Patricia B. Paz-Jacoba and Ricardo Jesus E. Gutierrez explain why the practical objective for a business should not be to eliminate the use of generative AI but to ensure that valuable AI-enabled outputs can be commercialized with a defensible chain of rights.

Businesses are already using generative artificial intelligence to produce advertising copy, product designs, software, reports, images, videos, music and other content. AI is also becoming part of research, product development and customer-facing services. The legal issue is therefore no longer simply whether a business may use AI. The harder questions are whether the resulting output can be owned, whether it can be commercialized safely, and what the business can do when others use AI to copy or misuse its content, brand or likeness.

Philippine intellectual property law still assumes human authors and inventors. Generative AI complicates that assumption because it can produce valuable material with little visible human input. It also makes copying faster and harder to trace. A business may spend heavily developing an AI-assisted asset, only to discover that its legal protection is uncertain and that imitations can spread across platforms within hours.

There is no single legal solution. An AI-enabled output may contain several distinct assets, each calling for a different form of protection.

Human authorship remains central to copyright

Under the Intellectual Property Code of the Philippines, copyright protection applies to original literary and artistic works. In February 2026, the Intellectual Property Office of the Philippines (IPOPHL) issued the Revised Rules and Regulations on Copyright Registration and Copyright Related Services (the 2026 Rules), which took effect on March 20, 2026.

The 2026 Rules are significant for businesses using generative AI. Rule I, Section 3.2 defines an author or creator as the natural person who created the work or any portion of it. Rule III, Section 1.2 recognizes authorship claims and disclaimers. Under Rule IV, Section 1, registration may be refused where the work lacks human authorship, originality or the minimum degree of creativity.

Accordingly, a work does not necessarily become protectable merely because a company paid for the AI tool, supplied prompts or caused the output to be generated. IPOPHL has previously taken the position that fully AI-generated works are not protected under the current copyright regime and that, for partially AI-generated works, protection extends to the portions created by natural persons.

The distinction matters in day-to-day practice. If a business uses AI to develop a campaign image, packaging design or video, it should establish the human-authored contribution to the final work. Drafts, source files, editing histories, prompt iterations, instructions to designers and records of human selection, arrangement or modification may all help. The more valuable the asset, the less sensible it is to wait for a dispute before assembling that record.

Ownership should also be addressed separately from authorship. Under the IP Code, copyright in a work created by an employee in the performance of regularly assigned duties generally belongs to the employer, unless there is an agreement to the contrary. By contrast, in a commissioned work, the person commissioning and paying for the work owns the work itself, but copyright remains with the creator unless there is a written stipulation to the contrary.

Businesses that use agencies, freelancers, consultants and AI-enabled creative vendors should therefore not assume that payment alone effects an assignment of copyright. Their contracts should clearly allocate ownership of human-created portions, rights in modifications and derivative works and responsibility for third-party materials incorporated into the output. The terms governing the AI tool should also be reviewed. A contractual provision stating that the user owns the output is useful between the user and the provider, but it cannot create statutory copyright where the output lacks the human authorship required by Philippine law.

Use the appropriate form of IP protection

The absence or uncertainty of copyright does not necessarily mean that an AI-enabled asset is commercially unprotectable.

For example, an AI-generated name, logo or visual device that functions as a source identifier may be capable of trademark registration if it satisfies the requirements of the IP Code. Trademark protection is particularly important for consumer-facing AI outputs because it protects the commercial identity of the brand rather than the authorship of the underlying artwork. A business should therefore consider trademark availability searches, clearance, and registration for any of its important AI-assisted marks, including brand elements, product names, logos and campaign identifiers, instead of relying exclusively on copyright.

For technical innovation, patent and utility model protection may also remain relevant. IPOPHL’s Examination Guidelines of Inventions, Utility Model, and Industrial Design Applications Relating to Artificial Intelligence that AI itself cannot be treated as an inventor. Nevertheless, AI may be used as a tool in the inventive process, and AI-assisted inventions may be patentable where a natural person is properly identified as the inventor, the invention makes the required technical contribution, and the applicable statutory requirements are satisfied. As discussed in the examination guidelines, the use of AI does not remove the requirements of novelty, inventive step, industrial applicability and eligibility as patentable subject matter.

Other commercially important assets may be better protected through confidentiality and contract restrictions. Proprietary datasets, prompt libraries, model configurations, fine-tuning methods, workflows and internal know-how may derive much of their value from not being publicly available. Businesses should consider access controls, confidentiality obligations, information classification and restrictions on uploading proprietary information into third-party AI systems.

In practice, an AI-enabled product may therefore require a combination of copyright, trademark, patent and contractual protection. Identifying the appropriate combination should form part of product development rather than be deferred until launch.

AI also changes the enforcement problem

Generative AI does not merely create new assets. It can also reproduce or imitate existing works, brands, voices and visual identities quickly and at low cost.

Copyright infringement remains relevant where AI-generated material reproduces a protected work or a substantial portion of it without authority. Section 216 of the IP Code also recognizes liability not only for direct infringement but, in specified circumstances, for a person who benefits from another’s infringing activity despite notice and the ability to control it, or who knowingly induces, causes or materially contributes to the infringement.

Those provisions may offer a starting point for disputes involving AI users, commercial deployers and intermediaries. Their application is not yet settled, however. Philippine jurisprudence has not defined the respective exposure of the person who entered the prompt, the business deploying the system, the model developer or the platform through which the output was distributed.

There is also an evidentiary problem. The right holder must still establish the factual basis for infringement, but generative AI systems may make it difficult to determine why a particular output resembles a protected work, what material was used to produce it or who caused the output to be generated. The model may be operated abroad, the relevant account may be anonymous and the content may be repeatedly altered and reposted. Accordingly, a commercially obvious instance of copying may still present difficult questions of attribution, evidence and jurisdiction.

Trademark and unfair competition laws may provide additional remedies when AI is used to imitate a business’s brand, trade dress or market identity in a manner likely to cause confusion or to pass off one’s goods or services as those of another. This is particularly relevant to AI-generated advertisements, fake endorsements, counterfeit listings and social media accounts that combine copied trademarks with synthetic content.

Deepfakes present a more difficult problem because the harm may extend beyond conventional IP rights. A synthetic video may copy copyrighted material, use a registered mark without registration, or falsely imply endorsement, but a person’s face, voice or general likeness does not necessarily fit neatly within the copyright or trademark framework. Depending on the circumstances, the Data Privacy Act, civil law remedies, cybercrime rules and other laws may also become relevant.

The regulatory framework is still developing. Several bills seeking to regulate deepfakes and digital likeness, such as House Bill No. 2312 (or the proposed Deepfake Regulation Act), House Bill No. 807 (or the proposed Take It Down Act) and Senate Bill No. 1714 (or the proposed Digital Likeness and Deepfake Regulation Act), are pending in Congress. The National Privacy Commission, in its Notice to the Public: Use of Real Persons’ Likeness in AI-Generated Images, has also emphasized the privacy risks arising from AI-generated imagery depicting identifiable persons without their consent. For businesses, this means that deepfake incidents should not be analyzed solely as copyright or trademark cases. The legal theory and enforcement route will depend on what was copied, how the synthetic content is being used, who is being harmed, and what remedy is required.

Takedown is often the first remedy, but not the last

For online infringement, speed may frequently be more material than a subsequent damages claim. A misleading AI-generated advertisement or deepfake can cause substantial commercial and reputational harm before a court case can be commenced.

Platform takedown procedures are therefore often the first enforcement tool. IPOPHL has also developed administrative mechanisms against online counterfeiting and piracy. Its Intellectual Property Rights Enforcement Office may take enforcement action in appropriate cases, and the Philippines has a voluntary administrative site-blocking mechanism under which copyright owners may seek the blocking of pirate websites.

These mechanisms, however, were principally designed for conventional counterfeiting and piracy. They may be less effective where infringing AI content appears as a single post, short-lived advertisement, anonymous account, or rapidly changing series of mirrors and reposts. Cross-border platforms also apply their own notice requirements and evidentiary standards.

Businesses should put an enforcement protocol in place before a serious incident. Preserve the evidence before sending a takedown request: the URL, account details, screenshots, timestamps, copies of the content, transaction records and any available metadata. The business should also be ready to prove its own rights through registrations, assignments, employment or commissioning agreements and records of human authorship.

Depending on the case, the appropriate response may involve platform notices, cease-and-desist letters, an IPOPHL administrative complaint, a site-blocking request, civil proceedings for injunction and damages or criminal enforcement. The correct sequence matters. An immediate takedown may stop the harm, but premature removal without preservation of evidence may make attribution and subsequent enforcement more difficult.  For example, if a platform removes a fake endorsement before the business captures the account details, timestamps and linked payment information, the evidence needed to identify the operator and trace related transactions may disappear with the post.

Building an enforceable AI portfolio

The practical objective for a business should not be to eliminate the use of generative AI. It should be to ensure that valuable AI-enabled outputs can be commercialized with a defensible chain of rights and that the business can act quickly if those rights are infringed or otherwise misappropriated.

This requires governance at the point of creation. At a minimum, businesses should:

• Approve the AI tools that may be used and the information that may be uploaded;
• Document human creative input and ownership of the resulting output;
• Identify when copyright or trademark registration should be pursued;
• Set the warranties and indemnities required from vendors and contractors; and
• Require legal and IP review before higher-value assets are publicly released.

The same discipline should apply to enforcement. A business that has preserved evidence of human authorship, secured written assignments, registered key marks and works where appropriate, and established a monitoring and takedown process will be in a materially stronger position than one that tries to reconstruct these matters only after an AI-driven infringement occurs.

Generative AI changes how quickly intellectual property can be created, copied, and distributed. It does not displace the fundamentals. Clear authorship records, ownership documents, appropriate registrations or recordations, provenance and an enforcement plan are now more important, not less. Businesses that address those issues while an AI-enabled asset is being developed will be better placed to protect and license it, and less likely to discover after launch that a valuable output is legally difficult to defend.