Colombia votes today in the presidential runoff, and with it comes the close of the most documented electoral artificial intelligence experiment in Latin America’s history. Polls are open from 8 a.m. to 4 p.m., and a result is expected tonight, with the Technology Ministry (MinTIC) and the ColCERT response team watching in real time, hour by hour (El Tiempo). We had been following this campaign for an uncomfortable reason: it accumulated six AI-made fabrications in just a few weeks and exposed a state capable of detecting them, but not of punishing them.
The gap is one of timing. Law 2502, the rule that provides the mechanism to sanction deepfakes, does not take effect until July 2026; until then, the electoral monitoring deployed on five simultaneous fronts can observe and warn, but cannot impose a single sanction. That is why the precedent Colombia leaves is ambiguous: it is not a regulatory success, but the most detailed map of the gaps the region will have to close before the next electoral cycle —Brazil votes in October 2026.
And there remains an underlying question that the day does not resolve: how much the deepfakes actually moved the result. There is still no clear methodology to measure it, so the experiment’s main unknown closes without an answer just as voting ends. Detecting disinformation without being able to sanction it does not stop democratic erosion: it documents it precisely.
In the region
The calendar does not stop at the vote. On June 25 and 26, four days later, the region’s technology ministers meet in Santo Domingo at the AI Ethics Forum for Latin America and the Caribbean, organized by UNESCO together with the development bank CAF and the Dominican government (Foro IA LAC). They arrive with the Colombian experience fresh and with a decision on the table: limit themselves to describing what happened, or commit to electoral frameworks that do have real sanctioning capacity before Brazil and the rest of the region vote. At the same time, Colombia also struck a more optimistic note: Uniminuto’s SAPIENS ecosystem, a network of AI agents for virtual education, was recognized at the Latam Digital 2026 Awards in the AI for social good category (Semana). The same country staging the region’s largest digital democracy experiment produces, on the same day, one of its first regional awards for AI applied to education.
Threads we’re following
The other story we had been following closely reaches a breaking point tomorrow. Nine days ago, the US government “switched off” Anthropic’s most powerful models —Fable 5 and Mythos 5— across the planet through an export control, after a way to get around their safety barriers was detected. The ban remains in place: Trump “softened” on June 20, according to public signals, but did not lift the directive (Anthropic). And starting June 23, access to Fable 5, which had been included at no cost in paid plans, will be charged as additional usage credits. It is the same lesson as all week, and it connects with Colombia more than it seems: access to the AI frontier —just like the capacity to sanction a deepfake— is being decided on state timetables, not market ones, and Latin America still has no seat at those tables.
If detecting disinformation without being able to sanction it only serves to document how democracy erodes, what would be the one concrete, enforceable instrument in 2026 that the region could bring to Santo Domingo to change the equation before the next elections?
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.