Recent developments of Artificial Intelligence and IP in Latin America (Pt. III)

As we speak, AI is spreading faster than electricity or the internet, with over 1.2 billion users globally. Companies are adopting AI to automate workflows, improve decision-making, reduce costs, and increase productivity. The global AI market is projected to exceed $4 trillion by 2030, driven largely by enterprise adoption (automation, coding, decision-making). 

To check our previous two articles: AI and Intellectual Property (Pt. 1): Safeguarding Creations in the Digital Age & AI and Copyright protection in Latin America (Pt. II).

“I saw it in the news.”

New WEF white paper on AI in Latin America

Recently, the World Economic Forum has published the white paper “Latin America in the Intelligent Age: A New Path for Growth”. Brazil, Chile, Mexico, Uruguay, and other Central American and Caribbean countries have engaged with UNESCO’s Readiness Assessment Methodology, with Paraguay following suit. Argentina, Brazil, Chile, Colombia, Costa Rica, Mexico and Peru have formally adopted the OECD’s AI Principles, which establish intergovernmental standards on AI. In 2025, Uruguay, a leader in developing regulation, became the first country in Latin America to sign the Council of Europe’s Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of Law, a legally binding treaty aimed at ensuring the responsible use of AI.

According to the White Paper, Latin America’s AI journey is characterized by rapid progress and persistent difficulties that need to be addressed in order to scale AI and fully realize its benefits. As we have shown, seven key challenges occur across the layers set out in the Blueprint: 1. Rising adoption has yet to deliver measurable impact; 2. Digital infrastructure divides remain acute, and resource needs, such as energy demand, can create new challenges; 3. Talent pipelines are weak; 4. Data readiness is lagging behind; 5. Governance and regulation are fragmented; 6. Capital is limited; 7. Regional collaboration remains an untapped opportunity.

Chile’s new open-source AI model is designed for Latin America

Latam-GPT is the result of a two-year regional effort led by the National Center of Artificial Intelligence of Chile, CENIA, and supported by over 30 institutions across eight Latin American countries. This language model was trained on the diverse cultures of Latin America, aiming to better reflect regional realities and strengthen the region’s presence in the global AI race.

Latam-GPT acts as a foundational infrastructure for future regional applications.

Country-by-country current legislative situation (May 2025)

Against that global backdrop, LATAM has moved from strategy statements to legislative drafting, with one country already across the line. 

Peru

Peru completed on 9 September 2025 a significant step in its AI framework by publishing the Regulation of Law No. 31814. The regulation, with a risk-based structure, sets out prohibited practices (including mass surveillance without a legal basis, crime prediction, and the use of autonomous lethal systems), and requires human oversight for high-risk uses. It also mandates coordination with the Peruvian National Institute for the Defense of Free Competition and the Protection of Intellectual Property (INDECOPI) and the National Data Protection Authority, creates regulatory sandboxes to support controlled trials, and establishes a phased implementation schedule with differentiated timelines for micro and small enterprises. Entry into force occurs after ninety business days, and the Secretariat for Government and Digital Transformation is tasked with issuing complementary rules and guidance. Taken together, these measures align Peru more closely with the global move toward use-based governance while preserving domestic institutional design.

Brazil

Brazil remains the region’s bellwether for a comprehensive, horizontal statute. The Bill PL 2.338/2023 sets out general rules for the development and use of AI, draws on a risk-based approach, and, in versions approved by the Senate on 10 December 2024, introduces transparency and supervisory architecture while signalling debates around training data disclosures and sectoral oversight. The bill is now in the Chamber of Deputies, which means the text can still change. Still, the direction is unmistakable: developers and deployers should expect duties to document, assess, and monitor high-impact uses, as well as to explain system behaviour to regulators and affected parties. This is not yet law; it is a legislative process with momentum.

Chile

Chile offers a government-sponsored bill that openly borrows the risk-based grammar while adapting it to local institutions. The executive introduced a draft law to regulate AI systems on 7 May 2024; it was unified with an earlier parliamentary initiative and entered its first constitutional stage in the Chamber’s Commission on Future, Science, and Technology. The public-facing materials make explicit what practitioners already see in the market: a classification of uses by risk, governance measures scaled to that risk, and a supervisory model tied to Chile’s forthcoming data protection authority. For EU SMEs, this means documentation and testing processes designed for the EU will read sensibly to Chilean reviewers, provided they are localised to the bill’s specific triggers and definitions as it advances.

Check our article on AI in Chile: GenAI: Balancing innovation and copyrights & the challenges of Chile’s AI Bill.

Mexico

Mexico’s Congress has been prolific in tabling initiatives. In 2024, a federal bill to regulate AI was introduced in the Senate with a structure that, again, mirrors the international risk-based approach and proposes an oversight authority. It is too early to treat any single text as the definitive Mexican framework. Still, the thrust is clear: lawmakers are converging on a taxonomy of unacceptable, high, and lower risk uses, combined with testing and registration mechanisms that will feel familiar to firms already building to EU-style expectations. Until a final statute emerges, SMEs should align operations with a management system approach and keep contractual documentation ready to show data provenance, model evaluation, and change control.

Colombia

Colombia has taken a different path by first adopting a national AI policy via CONPES 4144 in February 2025. Rather than imposing immediate horizontal obligations, the policy sets a government-wide program with actions and a budget through 2030 to build capacity, encourage adoption, and weave ethical guardrails into public and private use. Policies are not statutes, but they are not empty either: sectoral regulators read them as marching orders, and public buyers use them as templates for procurement. In practice, companies will find that adopting ISO/IEC 42001 or mapping to the NIST AI RMF earns credibility when engaging Colombian authorities during this implementation phase.

Argentina, finally, illustrates how multiple parliamentary tracks can move in parallel while the executive branch tests ideas in public debate. Alongside risk-tiered proposals introduced in 2023 & 2024, lawmakers in March 2024 tabled a high-profile bill colloquially known as the “Ley Turing,” which would, among other effects, reaffirm the role of human authorship for copyright and draw lines between AI-assisted and synthetic productions. Names aside, the harmony with peers is notable: legislators are borrowing the same organising concepts even as details vary. For an SME, that means a single internal governance stack (records, testing, transparency, provenance) will travel across borders, provided specific statutory clauses are checked country by country at launch.

IP protection strategies for EU SMEs

1) Treat patents and trade secrets as complements rather than substitutes. Novel architectures, training regimes tied to demonstrable performance gains, or concrete applications with measurable effect can make good patent cases if you are prepared to disclose enough to enable replication by a skilled person. Everything you do not need to publish to achieve that aim (i.e. the weights, pre- and post-processing heuristics, prompt libraries, failure taxonomies, and red-team playbooks) belongs behind your trade secret program, with real controls and audit trails. 

2) Code and documentation receive copyright protection. 

3) GUIs and distinctive visual elements may be eligible for design or copyright protection in some jurisdictions or unfair competition safeguards in others. 

4) Trade marks deserve early clearance and registration in key classes and markets. The trick is to decide up-front what you want to keep, what you want to share, and where each right delivers the highest return for the least disclosure.   

Contact us

If you have any doubts about how to benefit from IP protection in the AI sector and want to internationalize to Latin America, and know more about IP protection before you go, don’t hesitate to contact the Latin America IP SME HDs.