Today Colombia holds the first round of its presidential elections: 41.4 million eligible voters, 14 candidacies and, according to the Electoral Observation Mission (MOE), active circulation of candidate deepfakes, cloned audio and fake polls fabricated with artificial intelligence. All of this is happening without the country having an AI legal framework or specific electoral regulation for this type of content (NoticiasSuper).
The most talked-about case is a fake poll that showed one candidate winning and that circulated on WhatsApp: El Tiempo identified it as AI-generated because it still had the Gemini logo visible in the lower right corner. It is a revealing detail: the piece of disinformation did not even bother to remove the mark of the tool that created it. The MOE sums up the underlying shift this way: “the central challenge is no longer the physical tampering of votes, but the systematic deception of voters.” To gauge the terrain, the organization itself estimates that 78% of Colombians were exposed to digital disinformation in the past year.
The regional contrast is what makes this Sunday so interesting. Just four days earlier, Brazil approved its first comprehensive national AI framework, and its electoral court already operates with specific measures against synthetic content. Colombia goes to the polls with none of those instruments. If no candidate gets more than 50% plus one, there will be a runoff on June 21, and then the practical question will be whether it is possible to build any institutional response in just three weeks.
Also today
- OpenAI publishes its Frontier Governance Framework — It is the first public document in which the company explains how it aligns its safety practices with the European AI law and with California’s new transparency rule. It is voluntary and, for now, without independent external validation.
- OpenAI Codex comes to the Windows desktop — Its programming agent can now see, click and type in Windows applications, and be directed remotely from a phone. It is not available in the European Union, the United Kingdom or Switzerland.
- xAI opens Grok Build 0.1 on a public API — Another programming agent, this time with a very low entry price ($1 per million input tokens) and a score of 70.8% on the SWE-Bench Verified test, according to the company’s own measurements.
- Brazil bans AI “nudifiers” by decree — Two decrees signed in late May explicitly ban tools that generate fake nudity from real photos and require platforms to act proactively. It is the first country in the region with an active ban of this kind.
In the region
The week that is ending leaves Latin America with its densest regulatory map on AI to date, and at the same time exposes its unevenness. Brazil approved a national framework with a risk-based approach and fines that can reach 50 million reais, added decrees against “nudifiers” and on platform liability, and maintains specific electoral rules. Colombia, in the same week, votes with deepfakes circulating in real time and without an equivalent institutional tool. Big tech (Meta, OpenAI, Google, TikTok) has already voiced its rejection of the Brazilian decrees, citing “legal uncertainty,” a tension that will probably repeat itself when Chile, Argentina or other countries in the region move forward with similar rules. The debate, at bottom, is about who has the authority and the speed to respond when technology runs faster than the law.
Launches
- OpenAI Codex with desktop control on Windows — It is the first mass-use programming agent that leaves the isolated environment and operates directly on the real computer: it opens apps, clicks and types, and can be managed from a phone. Since Windows has high penetration in the region’s corporate PCs, the use case is quite direct for Latin American teams. For now it excludes Europe, the United Kingdom and Switzerland.
- xAI Grok Build 0.1 — A programming agent designed for multi-step tasks, with 256,000 tokens of context and a notably low price compared with comparable alternatives. For development teams looking for cheaper options, it is a new name to keep on the radar.
Threads we’re following
What is happening in Colombia adds to a story we have been following for weeks: that of a region regulating AI at two speeds. Brazil moved first with a general law and then with targeted decrees; meanwhile, other countries reach critical moments (such as a presidential election) without equivalent rules. In parallel, the race among programming agents keeps accelerating: within a month several tools capable of operating autonomously appeared, and each launch pushes prices down and capabilities up.
If Colombia ends up going to a runoff on June 21, are three weeks enough to improvise an institutional response to AI disinformation, or will the lessons only arrive in time for the next election?
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.