Living up to its name, the Government’s “miscellaneous” bill included, within its broad regulatory scope, a substantive limitation on intellectual property rights in the growing artificial intelligence business (a new article 71T for Law 17.336), which was rejected on Wednesday night during the project discussion held in the Finance Committee of the Chamber of Deputies.
Nothing new, as it literally transcribed the proposal from the Boric Administration, which had already been rejected just a few months earlier. What is notable, however, is that the very text serves to support diametrically opposed arguments: if before it was suggested that this would make Chilean-made artificial intelligence (AI) technologies flourish, for the current Government granting this generous guarantee to large tech companies — which, in essence, in Chile will never be sued by artists or publishers accusing unauthorized use of their works — would represent an important incentive for foreign investment in infrastructure and data centers in our country. An important and selfless contribution from the big tech companies for national reconstruction.
But beyond the initial embarrassment caused by the notorious legislative copy/paste, the worrying thing now is the stubborn attitude with which they intend to insist on this “open bar” for AI developers, since far from addressing the criticisms of the media, authors, and specialists, the substitute amendment presented last May 11 only worsened the problem: what was already an extraordinarily broad exception in favor of AI platforms was reformulated in even more indeterminate terms, including concepts such as “training, development or deployment” of AI models, up to “any other statistical analysis.” And the icing on the cake: the creation of a state fund managed by an interministerial committee, to “compensate” the holders of the works used to train the AI.
In other words, it is recognized that the massive exploitation of copyright-protected content does generate enormous profits, but the cost that AI companies — which will receive these economic benefits — would have to bear is transferred to the copyright holders exploited (and, incidentally, to the public treasury itself). Even more, the lucrative nature of these activities is made explicit, subjecting the state fund to the generation of gross income by AI companies, in clear contrast to activities framed within scientific or educational uses (which, needless to say, are already fully covered by the current Intellectual Property Law).
It is not, obviously, about questioning whether Chile should or should not join the global technological transformation represented by AI. The point is whether this should be financed through limiting the rights of authors, publishers, and media, or even by the Chilean State itself, instead of being borne by the companies developing this new and lucrative business. The previous Administration’s argument to incentivize a local industry to develop Chilean AI models also has little place: the project does not contemplate mandatory technology transfer, does not impose reinvestment obligations in Chile, does not establish limits on foreign exploitation of national content, nor anything of the sort.
The harm to the creative industries, on the other hand, is concrete and immediate. Publishers, musicians, journalists, photographers, designers, audiovisual producers, and creators of any kind of content whose economic activity depends precisely on licensing their works to those who use them, would have to tolerate that such protected content be massively used to train AI models, even capable of competing with the very industries that fueled their development.
Read more A federal judge orders a Colombian woman deported to Congo to be sent back to the United States
Because this so-called ‘intelligence’ also turns out not to be, after all, so artificial: beyond the famous Turing tests and philosophical discussions about the real nature of what we now know as generative AI, Large Language Models (LLMs) are not autonomous entities that create content out of nothing. The supposed ‘creation’ generated by these systems does not arise spontaneously from an independent intelligence, but constitutes the derivative result of human creation with which these models were trained. An AI engine cannot write a poem without having previously indexed thousands of works by human poets; it cannot create a song, a photograph, or a text without feeding on millions of works produced through the intellectual effort of real men and women. And it is this prior creative effort that this regulation ignores.
From a strictly legal standpoint, moreover, the proposal conflicts with international copyright law: exceptions to intellectual property are subject to the so-called “three-step test” of the Berne Convention, which requires that any exception (i) be limited to specific cases; (ii) not affect the normal exploitation of the work; and (iii) not cause unjustified harm to the legitimate interests of the rights holder. Thus, the phrase contained in this new negative amendment about avoiding a “conflict with the normal exploitation of the work” is almost ironic, if in practice rights holders would lack mechanisms to even identify whether their works were used to train AI models or not, since the project also contains no kind of transparency or accountability obligation to that effect.
It is not arbitrary to ask, then, whether an exception of this breadth could also entail constitutional flaws by compromising the protection that the Fundamental Charter guarantees over intellectual creations: if copyright precisely means the exclusive power to authorize or prohibit the economic use of a work, a rule that deprives ex ante rights holders of any possibility of authorization or remuneration for massive and lucrative technological exploitations would effectively affect the content or essential core of such constitutional guarantee.
How then should the economic benefits of AI be distributed between tech companies and copyright holders? It seems we have seen this movie before, haven’t we? It happened at the end of the last century, with digitization and platforms like Napster or Cuevana, when it was argued that the internet had “killed” copyright and that music, movies, and content would now all be free. History proved exactly the opposite: technology changed, business models evolved, and copyright adapted. Streaming platforms, mass licensing systems, and monetization mechanisms compatible with the digital environment emerged.
The contemporary challenge, and Chile’s challenge, is not to choose between artificial intelligence and intellectual property, but to build a balanced innovation model that allows technological development without turning authors and creators into involuntary and unpaid suppliers of inputs for global tech monopolies. Intellectual property, although some may not like it, has not died with AI. On the contrary, AI feeds on it and grows thanks to it. And there lies the paradox, for nothing could be more inhuman than to disregard the creative effort of flesh-and-blood people on which artificial intelligence is based and developed.
Read more Sheinbaum invites Ayuso to spend more vacation time in Mexico: “She would learn a lot”
