UNESCO’s Third Ministerial Summit on the Ethics of Artificial Intelligence in Latin America and the Caribbean is now in session: it opened today in Santo Domingo, and the region’s technology ministers have just 48 hours to produce a declaration, scheduled for June 26 (UNESCO). What is new compared with previous days is not that the summit is taking place —that was already expected— but the case file that arrives closed at the table: this month Colombia certified a presidential result with digital interference documented by the state itself, and the US ban on the Fable 5 model entered its third week without the region having been consulted. Neither of those two facts was on the original agenda; both now define the forum’s underlying question.
That question is not whether there will be a declaration, but whether the document will name what the region lived through: a unilateral “off switch” on frontier models with no regional consultation mechanism, and an election with digital interference in the official record. On June 24, Colombia’s National Electoral Council certified Abelardo de la Espriella as president-elect with 49.66% of the vote, by a margin of less than one point (Registraduría). It is the region’s first final presidential result that coexists with state documentation of AI interference during the vote —not as a candidate’s complaint, but as a finding by the ICT Ministry itself. The detection mechanism worked; the sanctioning one was not activated, and that gap between detecting in real time and being unable to sanction in real time is exactly what the summit could try to regulate, with Brazil voting in October.
The corporate backdrop also shifted. Anthropic replaced its chief executive, Dario Amodei, with researcher Tom Brown as its main point of contact with the US agencies negotiating the lifting of the ban, a change that prediction markets read as positive: the probability of resolution before July 1 rose from 52% to 60% (Anthropic). The gesture signals a shift in the negotiation from the political to the technical —Brown can demonstrate how the model’s controlled environment operates—, and it comes after Congress put the spotlight on the conflict of interest of negotiating the ban while the company had an IPO in process.
Also today
- Colombia certifies De la Espriella as president-elect by less than one point — The final result coexists with digital interference recorded in the state’s official record; the threshold for annulment was not reached. The first legal precedent of its kind in the region ahead of Brazil’s elections.
- Gemini 3.5 Pro is officially postponed to July — Google confirmed that the model will not reach general availability within the June window. With Android at more than 80% penetration in the region, the delay also postpones its default deployment on hundreds of millions of Latin American devices.
- Colorado weakens its high-risk AI law under industry pressure — The state replaced SB-205 —the first US high-risk AI framework, from 2024— with a narrower version. The standard that part of the region looked to as a reference is being weakened in the same week its ministers are discussing governance in Santo Domingo.
OpenAI also added its first in-house inference chip, developed with Broadcom to lower the cost of running its models and reduce its dependence on NVIDIA; if the bet scales, it changes the economics for labs that do not yet have their own silicon —Anthropic among them, which keeps running Fable 5 on other companies’ hardware.
In the region
Santo Domingo is the first test of whether Latin American multilateralism can produce concrete instruments and not just declarations of principles. The Évian G7 closed days ago without a common AI framework, the Fable 5 ban has been in place for almost two weeks without regional consultation and Colombia certified a result with digital interference in the official record: three facts the ministers cannot reverse, but can use to set what minimum instrument —enforceable and verifiable— the region would commit to having operational before October. That conversation about the region’s own capabilities is the context for the Guía País IA 2026 that the Chilean association ACHIADS will present on June 30, the first systematic map of Chile’s AI capabilities by sector.
Launches
- GPT-5.5-Cyber, full version via OpenAI Daybreak — One of the first frontier autonomous cybersecurity capabilities accessible in the region during the Fable 5 ban. It has three access tiers; the general tier opens with a standard account, while the full offensive capabilities require a contract with certified partners. IBM, which distributes it, has an active presence in Brazil, Mexico, Colombia, Argentina and Chile.
Threads we’re following
The US ban on Fable 5 —which we have been following since it was announced on June 12— reaches its thirteenth day with the most concrete sign yet of a possible way out. The replacement of Dario Amodei by Tom Brown at the head of the negotiation, together with the jump in prediction markets to 60%, suggests that the conversation has stopped being about permissions and has become about verifiable technical conditions. For the region, the underlying question remains open and is the same one sitting at the table in Santo Domingo today: if the ban is lifted with conditions negotiated with a specific lab, did Latin America have any voice in defining which conditions are considered sufficient? (status of the ban).
If the June 26 declaration could contain a single concrete instrument —enforceable and verifiable before October 2026— that responded both to access to frontier models without regional consultation and to the detection of electoral interference without sanctions, what would it be, who would carry it out and what would make it binding?
Correction (September 30, 2026). The original version named Colombia’s president-elect as “Nicolás De la Espriella”; the correct name is Abelardo de la Espriella, according to El Tiempo.
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.