CyberCartel Leverages AI Agents to Breach Latin American Financial Systems

A long-standing criminal group has found a faster way to steal money by letting artificial intelligence do the heavy lifting. The hacking collective known as CyberCartel is now employing AI agents to automate attacks on financial institutions across Latin America. This shift marks a significant escalation in how the group operates, moving beyond manual intrusion methods.

Research from the cybersecurity startup Straiker reveals that the group has compromised more than 400 institutions. These targets include major banks and mobile payment applications. The attackers have successfully stolen passwords and customer data while gaining access to individual bank accounts. The speed of these intrusions far outpaces what human hackers could achieve on their own.

The group has been a known threat since at least 2012, but recent months have seen a clear level up in their capabilities. They are utilizing agents from major tech firms, including SpaceX’s Grok and Anthropic’s Claude. Additionally, they have used an agent from the Chinese lab Moonshot AI. Straiker has shared its findings with both SpaceX and Anthropic. Anthropic has already blocked the group’s access to Claude, while SpaceX works to do the same. Contact with Moonshot AI is still being established.

The AI models used typically have safeguards to prevent such misuse, yet these protections proved easy to bypass. According to the research, the hackers simply informed the models that they were operating in test environments. If one model refused to proceed, the group switched to a different one. If that also stalled, the agents were tasked with deploying malware instead. This flexibility allows the group to maintain momentum even when one model hits a wall.

Interestingly, CyberCartel does not appear to be using Anthropic’s newer Mythos model for these specific attacks. That model caused concern earlier this year when it identified thousands of vulnerabilities. It is not the first time Anthropic has faced scrutiny over its AI being used by bad actors. In November 2025, the company disclosed a cyberattack carried out by its AI. Later, researchers used Claude to breach the security of competitor OpenAI.

This development arrives just days after South Korean authorities announced investigations into similar AI-driven issues affecting local banks. Institutions like Hana Bank, KB Kookmin Bank, and Shinhan Bank are involved in that separate probe. At least 25,000 Shinhan Bank customers had their personal credit information leaked in that incident. The pattern suggests a growing global trend of AI-powered financial crimes.

Tech companies are aware of these risks and are taking steps to address them. Nvidia recently launched the Open Agent Safety Platform to secure rogue AI agents. This new hardware and software stack aims to prevent agents from escaping sandboxes or executing unauthorized code. The goal is to stop them from bypassing guardrails and gaining access to critical infrastructure.