The Government of President José Antonio Kast submitted to Congress, on April 22, 2026, the “Bill for National Reconstruction and Economic and Social Development” (Bill No. 18216-05, Presidential Message No. 018-374). Article 8 thereof introduces a new Article 71 T into Law No. 17,336, allowing—without authorization or payment to the right holder—the reproduction, adaptation, distribution, and public communication of lawfully published works when the use is aimed at “the extraction, comparison, classification or any other statistical analysis of language, sound or image data.”
This rule effectively enables the large-scale training of artificial intelligence models using protected Chilean content. The proposal literally replicates the wording previously introduced by the prior administration (Boric government) in Bill 16821-19 (May 2024), which was rejected by the Chamber of Deputies both in the Future, Science and Technology Committee and in plenary session in October 2025.
The provision has triggered broad opposition from SCD, ANATEL, ANP, UNA, SECH, and ARCHI, as well as concerns within the ruling coalition itself. For right holders and AI developers exposed to the Chilean market, the discussion is critical: it will determine whether Chile adopts a text and data mining (TDM) exception more permissive than the European system (i.e., without opt-out, without distinction between scientific and commercial use, and without dataset transparency obligations), or maintains the current framework, where AI training on protected works generally requires a license.
What does the proposed Article 71 T say?
The literal wording of Article 8 of Bill 18216-05 is as follows:
“Article 71 T.—It shall be lawful, without remuneration or authorization from the right holder, to perform any act of reproduction, adaptation, distribution, or communication to the public of a lawfully published work, where such act is carried out exclusively for the purpose of extraction, comparison, classification, or any other statistical analysis of language, sound, or image data, or other elements composing a large number of works or a large volume of data, provided that such use does not constitute a disguised exploitation of the work or protected works.”
The drafting contains three decisive features for risk analysis:
First, it covers the four core economic rights (reproduction, adaptation, distribution, and public communication), thus encompassing the entirety of economic exploitation rights recognized under Chilean law.
Second, it does not include a rights reservation mechanism (opt-out), nor does it distinguish between non-profit scientific use and commercial exploitation. Unlike the European model (Directive (EU) 2019/790, Articles 3 and 4), the exception operates automatically and forces the right holder to prove, ex post, that the use constitutes “disguised exploitation.”
Third, the “disguised exploitation” clause is the only limiting factor, making it vague and arguably poorly drafted, with the burden of proof effectively shifted onto the affected author.
The bill was submitted with “urgent” status and referred to the Finance Committee of the Chamber of Deputies (rather than to Culture or Future and Science), a procedural issue criticized by industry associations.
Current Chilean law vs. the proposed regime
As of April 24, 2026, Chilean law does not include any exception for TDM or AI training. Law No. 17,336 provides a closed list of limitations (Articles 71 A to 71 S), all subject to the three-step test (Article 71 A): special cases, no conflict with normal exploitation, and no unreasonable prejudice to the right holder.
None of these exceptions, under prevailing doctrine, allows large-scale commercial AI training. As a result, developers training models on protected Chilean works without a license operate in a prima facie infringing context.
The European model: the benchmark Chile is not following
Directive (EU) 2019/790 established a dual regime:
- Article 3: mandatory TDM exception for research organizations and cultural heritage institutions.
- Article 4: general TDM exception, subject to lawful access and absence of an opt-out.
This framework was reinforced by the AI Act, which:
- Requires providers of general-purpose AI to respect rights reservations (Article 53(1)(c)).
- Imposes transparency obligations regarding training datasets (Article 53(1)(d)).
- Applies extraterritorially to any model placed on the EU market.
Non-compliance may result in fines of up to 3% of global turnover or €15 million.
In contrast, the Chilean proposal includes none of these safeguards.
The U.S. approach: fair use under pressure
The United States relies on the fair use doctrine (17 U.S.C. § 107), applied case-by-case. Recent case law shows increasing scrutiny of:
- Dataset provenance
- Market substitution effects
- Commercial exploitation
While some decisions have favored AI developers, the legal landscape remains fragmented and uncertain.
International constraints on Chile
Chile is bound by international treaties such as the Berne Convention, TRIPS, and WIPO treaties. Any new exception must comply with the three-step test:
- Special cases
- No conflict with normal exploitation
- No unreasonable prejudice to right holders
The breadth of Article 71 T raises serious doubts as to its compliance with these obligations.
Practical implications
For right holders:
- Risk of uncompensated use of protected works
- Need for defensive strategies (opt-outs, licensing, litigation readiness)
For AI developers:
- Continued infringement exposure under current law
- Extraterritorial compliance obligations (especially in the EU)
- Increasing due diligence expectations (dataset traceability, licensing, filtering)
Conclusion
The bill represents Chile’s second attempt in under two years to introduce a highly permissive TDM exception. The global trend—particularly in Europe and increasingly in the U.S.—is moving toward structured licensing, dataset transparency, and enforceable opt-out mechanisms.
Chile now faces a strategic regulatory choice: align with this international trajectory or adopt a markedly more permissive model with uncertain legal and economic consequences.