This week Chile’s Senate passed its economic recovery “mega-reform” without the article that would have exempted artificial intelligence systems from paying copyright fees for using protected works for training. The provision, already rejected twice before under pressure from publishers and the audiovisual sector, was left out of the final text after a marathon session (Bío Bío Chile). At first glance, it looks like a point for creators. But on closer inspection, the debate did not end: it moved to a different legislative vehicle.
Coverage in recent days revealed a finer detail. The AI bill still before the Senate (Boletín 16.821-19) includes a data mining exception almost identical to the one that fell out of the mega-reform. The key difference is that this version does include an opt-out mechanism: rights holders can expressly reserve their rights and be excluded from training. It is a safeguard the rejected article never offered. In other words, the underlying dispute between those who create content and those who develop AI did not disappear with the rejection; it moved to the law designed specifically to resolve it, and this time with an exit door for authors.
The move matters beyond Chile. It is one of the first cases in the region where the question (does AI training need authors’ consent?) is being settled piece by piece, through trial and error, rather than all at once. That the second version comes with an opt-out suggests the system learned something from the failed attempt: it is not enough to allow the use of protected works; the people who created them need a verifiable way to say no.
Also today
- Anthropic schedules investor meetings for a possible IPO in October — the company already trades on the secondary market above OpenAI, with an estimated valuation between $1.05 trillion and $1.2 trillion.
- Meta employees sue the company for using AI to decide layoffs — 26 workers allege a disproportionate impact on people on medical or parental leave, among the 8,000 laid off.
- OpenAI answers Apple’s lawsuit by blaming a name mix-up — it says it has no evidence of trade secret theft and attributes its silence to an internal error by its legal team.
- With AI and nuclear physics, a papyrus carbonized by Vesuvius is read for the first time — the text turned out to be a treatise on Stoic ethics from the 2nd century BC, recovered without physically unrolling the scroll.
- AI services offer to create “digital ghosts” of deceased people — conversational replicas that reopen questions about grief and posthumous consent.
In the region
Beyond Chile, the week brings two governance moves. In Argentina, Decree 581/2026 transferred political control of the AAIP (the personal data protection authority) and of ENACOM from the Secretariat of Science and Technology to the Chief of Cabinet’s Office (Infobae). It is an internal reshuffling of power that, in practice, recentralizes the body that regulates data just as its role in AI governance becomes more relevant, reducing its technical autonomy. In Mexico, meanwhile, the expectation remains that the formal legislative debate on AI will begin in the Chamber of Deputies on July 19.
Launches
- Inkling — the first open-weights model from Thinking Machines Lab, Mira Murati’s lab. It is a mixture-of-experts architecture (which activates only part of its parameters for each query) with 975 billion parameters, 41 billion of them active, multimodal and with a one-million-token context. It is downloadable and customizable, with no closed licenses.
- Grok Build, now open source — xAI released the full code of its agentic coding tool under the Apache 2.0 license, after the scandal over a repository leak. Now anyone can audit it or run it locally.
Threads we’re following
This is not the first time Chile has run into the question of copyright and AI. In recent weeks we had been following how the country debated whether training models on protected works should require paying their authors, and how the Finance Ministry had ruled out a broad exemption. Today’s chapter closes that attempt via the mega-reform, but opens another: the real debate now plays out in the dedicated AI law, and the appearance of an opt-out for rights holders is the new element that was not on the table before.
If Chile is struggling to decide once and for all whether AI training on protected works needs authors’ permission, and ends up settling it piece by piece, how ready is the region for the harder questions already looming, such as whether an AI can decide who gets laid off without verifiable human oversight?
About this entry. It is generated automatically from public sources, without human review before publication. It may contain errors of interpretation or summary; please check each story against its original source (the links lead there) before citing it or making decisions based on it.
Doble Click is written with Anthropic models.