AI and Intellectual Property: Guide for China, India, LATAM, and South-East Asia

As EU companies deploy artificial intelligence in product development, creative production, and research, the question of how to protect AI-generated outputs and AI-enabled inventions across global markets has become a core compliance challenge. The legal position on AI and intellectual property varies substantially between China, India, Latin America, and South-East Asia — and diverges significantly from the European framework. This guide provides a practical overview for EU companies seeking to structure their IP strategies across these regions.

The EU Baseline: Text and Data Mining, Platform Liability, and AI-Generated Works

Under the EU Copyright in the Digital Single Market Directive (Directive 2019/790), Article 4 establishes a broad text and data mining (TDM) exception for commercial purposes, subject to rights-holders' opt-out rights. Article 17 addresses platform liability for user-uploaded content, introducing upload-filter obligations for large platforms. These provisions shape the landscape in which EU companies train AI models and deploy AI-generated content.

The EU has not enacted specific legislation recognising AI-generated works as copyrightable in the absence of human authorship. The European Court of Justice's line of authority on originality — requiring a work to be the author's own intellectual creation — means that purely AI-generated outputs currently lack copyright protection in the EU unless a human's creative choices are sufficiently reflected in the result.

China: Emerging Protection for AI-Generated Outputs

China has moved faster than most jurisdictions in addressing AI-generated content. In 2023 and 2024, Chinese courts issued rulings in cases involving AI-generated images and text, holding that content produced with meaningful human creative input can qualify for copyright protection under the Copyright Law of the People's Republic of China. The key principle is that the human using the AI tool must exercise sufficient creative control — through prompting, selection, or modification — to be recognised as the author.

For AI inventions, China Patent Law Article 2 defines an invention as a new technical solution relating to a product, process, or improvement. Chinese patent practice requires a human inventor to be named. However, the State Intellectual Property Office (CNIPA) has issued examination guidelines indicating that inventions produced with AI assistance, where a human directed the inventive process, can qualify for patent protection. The decisive question is whether a natural person made the inventive contribution.

For EU companies, China's approach means that AI-generated content can attract IP protection in China if the human-input element is documented at creation. Prompt engineering decisions, editorial selections, and iterative human feedback should be recorded as part of the IP creation record to support potential Chinese copyright claims.

Article reference: China Patent Law Article 2; China Copyright Law Articles 2 and 11.

India: Conservative Patent Practice, Evolving Copyright

India's approach to AI and patents remains conservative. The Patents Act 1970, as amended, excludes abstract mathematical methods and mental acts from patentability, and Section 3(k) explicitly excludes computer programs per se. The Indian Patent Office applies a strict test: an AI-related invention must produce a "technical effect" beyond the computer program itself, and the specification must demonstrate this technical character. AI inventions that can be framed as producing a concrete technical result — improved hardware performance, novel sensor data processing, physical manufacturing outcomes — have a better chance of success.

On copyright, India's Copyright Act 1957 Section 2(d) defines "author" for computer-generated works as the person who caused the work to be created. This provision, originally drafted for earlier automation contexts, has been interpreted by Indian courts and the Copyright Office as capable of covering AI-generated works where a human operator is identifiable as the directing cause. This creates a more favourable default position for AI-generated outputs in India than in the EU.

EU companies should document the human decisions that initiated and directed AI output generation in India-facing projects, ensuring that a named natural or legal person can be identified as the "person who caused the work to be created" within the meaning of the Indian Copyright Act.

Latin America: Fragmented National Laws

Latin America does not have a unified supranational IP framework comparable to the EU. The Andean Community (CAN) Decision 486 on industrial property and Decision 351 on copyright apply in Bolivia, Colombia, Ecuador, and Peru, but other major economies — Brazil, Mexico, Argentina, Chile — operate under their own national IP laws.

On copyright, most Latin American jurisdictions follow the Berne Convention's human authorship requirement. Brazil's Lei de Direitos Autorais (Law 9.610/1998) and Mexico's Ley Federal del Derecho de Autor require a human creator. AI-generated works without identifiable human creative input generally cannot be registered or enforced in these jurisdictions.

On patents, Brazil's INPI and Mexico's IMPI apply examination standards comparable to the European Patent Office on AI subject matter: software-implemented inventions must produce a technical effect beyond normal computer interaction. Article 52 EPC (European Patent Convention), which excludes mathematical methods and programs as such, is often used as a reference standard by Latin American patent offices when developing their own guidelines, giving EU companies some familiarity with the conceptual framework even where national rules diverge in detail.

Trade secret protection is generally stronger than patent or copyright for AI assets in Latin America, particularly in Brazil, where Law 9.279/1996 provides robust trade secret remedies including criminal sanctions for misappropriation.

South-East Asia: Varied Positions, Active Reform

South-East Asia presents a highly varied landscape. Singapore has the most developed AI IP framework in the region, with the Intellectual Property Office of Singapore (IPOS) having issued guidelines on AI-generated works and AI-assisted inventions. Singapore's Copyright Act 2021 does not provide copyright to AI-generated works absent human authorship, but Singapore patent practice — modelled on UK and EPO principles — is receptive to AI-assisted inventions where human inventive contribution is documented.

Thailand, Indonesia, Malaysia, and the Philippines are at earlier stages of AI IP policy development. All are Berne Convention members, meaning that copyright protection is available for works with human authorship, but none has enacted specific AI-generated work provisions. Patent laws in these jurisdictions follow TRIPS Agreement requirements and generally exclude software and AI methods as such.

For EU companies with South-East Asian operations, trade secrets and contractual IP protections offer more reliable near-term protection than patents or copyright for AI-generated assets, pending further legislative development in the region.

Practical Implications for EU Companies

EU companies expanding globally should implement a layered IP strategy that combines human authorship documentation (for copyright), inventive contribution records (for patents), and robust trade secret protocols (for algorithm protection). The lack of harmonisation across China, India, Latin America, and South-East Asia means that a single IP posture will not suffice: regional counsel and jurisdiction-specific filings remain necessary.

The EU's own TDM exception under Directive 2019/790 Article 4 should be factored into training data decisions for models deployed globally, as the absence of equivalent TDM exceptions in many jurisdictions outside the EU creates potential copyright infringement exposure for training data use.

This article is informational only and does not constitute legal advice. Consult qualified legal counsel for advice specific to your organisation.

Frequently Asked Questions

Can an EU company register copyright in AI-generated content in China?

Yes, provided a human's creative contribution can be demonstrated. Chinese courts have recognised copyright in AI-assisted works where the human user exercised meaningful creative control through prompting, selection, or editing. EU companies should maintain records of human creative decisions made during AI-assisted content production to support registration and enforcement in China.

Does Article 52 EPC apply to patent filings in Latin American countries?

Article 52 EPC applies directly only to European Patent Convention member states and regional filings before the EPO. However, several Latin American patent offices use EPO examination guidelines as a reference when assessing AI-related inventions, making EPO case law relevant as persuasive authority even where it is not formally binding.

Is trade secret protection reliable for AI algorithms in South-East Asia?

Trade secret protection varies in quality across South-East Asia. Singapore offers the strongest statutory and common law protection. Thailand, Indonesia, and Malaysia have trade secret laws but enforcement infrastructure is less developed. Companies should combine contractual confidentiality obligations, technical access controls, and — where available — trade secret registration or notification procedures with competent authorities to maximise protection.

Sources

• EU Copyright in the Digital Single Market Directive 2019/790 — EUR-Lex
• CNIPA — Guidelines for Patent Examination, AI-Related Inventions — China National Intellectual Property Administration
• IPOS — Intellectual Property and Artificial Intelligence — Intellectual Property Office of Singapore