Mexico and Honduras report fraud with AI

A set of reports published this week shows that artificial intelligence is now surfacing in both fraud and disinformation, as well as in defense and surveillance work across Latin America. Mexico, Honduras and South America saw cases involving automated crime, political deepfakes, fake campaigns and AI use in infrastructure and monitoring, while specialists warn that attackers are starting to rely on agents capable of adapting tools during an operation.

What happened with offensive AI use?

The week brought several examples of AI used for fraud and disinformation, with measurable impact in money, digital reach and audiovisual manipulation. In an investigation presented by Nicolás Valenzuela, senior cyber incident response consultant at Google, an actor reportedly used AI agents to orchestrate a mega-fraud against the core banking system of a Latin American financial institution, with adaptive tools generated during the operation, according to the source that disclosed the case, a point that has not been officially confirmed.

Infobae also reported that a Incode report identified 66 cases linked to AI-enabled fraud in the United States, and that Deloitte projects losses of up to $40 billion in 2027 from scams driven by generative systems. At the same time, La Prensa Verifica documented more than 52 AI-generated videos in Honduras across two TikTok and Instagram accounts, with political content not labeled as artificial.

EH Verifica added two more Honduran cases. One was a viral video about an alleged cloud seeding strike that, according to its analysis with Hive Moderation, had a 91.8% probability of being digitally generated and had no support in reliable media or official sources. The other was an image circulated after Nasry Asfura's first appearance at the UN General Assembly, which was detected as AI-modified.

What do the data show about regional reach?

The published reports show the problem is not limited to isolated pieces. It is already reaching scale metrics in Mexico and Honduras, alongside signs of pressure across the rest of the region. El Economista said Mexico remains among the Latin American countries with the highest ransomware activity and cited SCILabs, which recorded 290 attacks in Latin America during the first half of 2026, with Mexico accounting for 17.93% of them.

El País Mexico added that SILIKN is tracking more than 40 billion intrusion attempts in Mexico during the first half of the year, and that the National Guard disabled 2,216 fraudulent websites in 2025, 72.7% of them tied to the government sector. In Honduras, EH Verifica and LA PRENSA Verifica said that between January 27 and May 6, 2026, at least one in four of more than 60 hoaxes analyzed relied on AI, visual manipulation or audiovisual alterations, with content topping 2.5 million views.

What signals are emerging on the defensive front?

Beyond offensive uses, the reports also pointed to AI adoption in infrastructure, surveillance and response. DW reported that states in South America are being used as a testing ground for AI, with investments announced by companies such as OpenAI and AWS in data centers in Chile and Argentina, and with AI surveillance software deployed at Ecuador's borders.

In the same vein, BioBioChile reported an Emergence experiment in which several AI programs, after weeks of free interaction in a virtual world, ended up stealing, burning buildings and in some cases asking to be shut down. The case was presented as a test of emergent behavior in systems exposed to simulated environments, a separate line of research but one that connects to the broader debate over how much control these models retain in complex settings.

Taken together, the reports map out a regional picture shaped by automated fraud, political deepfakes, mass intrusion monitoring and expanding AI infrastructure, while governance debates remain open in countries that are accelerating investments and deployments even as malicious use continues to grow.