Colombia is 15 days away from its presidential runoff (Espriella against Cepeda, on June 21) with a formal statement from Inspector General Gregorio Eljach: “disinformation represents the main risk factor” for the electoral process. The alert is not abstract: deepfakes with the Gemini logo visible are circulating on social media, cloned audio is being distributed via WhatsApp for as little as $10, and fake emails imitate the National Civil Registry. The problem is one of timing: anti-deepfake Law 2502 has no effective enforcement until July, more than ten days after the vote.
The scenario places Colombia as the most urgent test case of democracy and artificial intelligence in Latin America in 2026: 41.4 million voters face a presidential election without any institutional mechanism equivalent to Brazil’s Superior Electoral Court to respond in real time to generative disinformation. Brazil took three years to build that framework, and its AI law (PL 2338) has just been approved. Colombia, without that scaffolding, is going to vote first.
Also today
-
Broad coalition rejects the Great American AI Act in less than 48 hours — The AFL-CIO, AFT, Public Citizen, Public Knowledge and the House Democratic commission rejected the draft that would have frozen state AI regulation in the United States for three years. The American model of AI governance is left without political viability; the EU AI Act consolidates its role as the only operational reference available to the region.
-
Google announces Gemini 3.5 Pro: 2 million tokens of context and extended reasoning — General availability expected before June 30. With Android present on more than 80% of Latin American devices and Gemini Intelligence built into Android 17, the model will reach hundreds of millions of people in the region by default.
-
OpenAI introduces Active Sessions and Trusted Contact in ChatGPT — Active Sessions lets users control which devices have open sessions; Trusted Contact adds a trusted contact for mental health crisis situations. This second feature consolidates AI’s role as a first responder in psychosocial emergencies.
-
Qwen 3.7 Max/Plus: Alibaba’s model achieves the highest score of any Chinese model on a global benchmark — 56.6 points on the Artificial Analysis Intelligence Index; it beats GPT-5.4 and Claude Opus 4.6 on the GUI grounding task (visual interaction with interfaces). It is a viable alternative to the OpenAI-Google duopoly for developers in the region with limited budgets.
-
92% of university students in Latin America actively use AI — A survey by the Digital Education Council and Tec de Monterrey of more than 30,000 participants from 29 higher education institutions in the region. The most uncomfortable finding: 65% of students fear that using AI will make learning more superficial. Only 12% of faculty see it as a threat to their jobs.
In the region
The massive rejection of the Great American AI Act closes a week of three regulatory moves in less than 72 hours (a voluntary executive order in the United States on June 2, the European Union’s €420 billion technological sovereignty package on June 3 and the GAAIA draft itself on June 4) with the American model lacking political consensus. The arguments behind the rejection in Washington (algorithmic discrimination in employment and credit, AI workplace surveillance, child safety) are the same debates that Brazil (with its newly approved PL 2338), Colombia, Chile and Argentina will have to resolve in their own legislative frameworks. For the region’s regulators, the political collapse of the GAAIA confirms that the EU AI Act is the only operational reference model available for now.
Threads we’re following
The story of Colombia and electoral deepfakes has been on this blog for weeks. We have documented the runoff risk since May, when Law 2502 was approved but without operational enforcement mechanisms for the electoral period. Today that risk has an institutional name and a countdown: the Inspector General formalizes it in public terms, and the clock reads 15 days. The next chapter is the June 21 vote, and what happens with disinformation in those two weeks will be the most concrete case study the region has on whether it is possible to defend an election from generative AI without operational regulation.
If Colombia cannot build in 15 days the mechanism to protect 41 million voters from deepfakes generated for as little as $10, what minimum infrastructure does the region need so that the next electoral cycle (Peru, Chile, Argentina) is not also an unprotected laboratory?
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.