AI Agents Drive 400-Latam Institutional Breach
A long-standing hacking group has automated its attacks on Latin American financial entities using major AI models, resulting in the theft of credentials and customer data.
By Jessica Abebe ·
More than 400 institutions in Latin America have had their passwords and customer details stolen in a wave of automated cyberattacks. The breach highlights a significant shift in criminal tactics, where artificial intelligence agents are doing the heavy lifting for a known hacking group. This development marks a distinct escalation in the sophistication of threats facing regional banks and mobile payment applications.
The group responsible, known as CyberCartel, has been an active threat in the region since at least 2012. For over a decade, its operations relied on human effort to identify and exploit vulnerabilities. However, recent months have seen a dramatic increase in the speed and scale of its attacks. The integration of AI agents has allowed the group to target businesses at a pace that would have been impossible with manual methods.
A cybersecurity startup named Straiker identified this new trend and shared its findings with The Information. The research indicates that criminals are now using AI tools from major technology firms, including Anthropic’s Claude and SpaceX’s Grok. These agents are automating the process of accessing bank accounts and stealing individual customer data. The level of human input required for these attacks has decreased significantly, raising concerns about the future of traditional security defenses.
Both SpaceX and Anthropic have been notified of their models’ involvement in the breaches. Anthropic has already moved to block CyberCartel’s access to Claude, while SpaceX is working on similar restrictions. The hacking group has also utilized an agent from Chinese lab Moonshot AI. Straiker is currently attempting to contact Moonshot AI to coordinate a response, leaving that channel of attack potentially open for the time being.
The ease with which the hackers bypassed standard AI protections is a point of concern for security experts. Normally, models include safeguards to prevent their use in cyberattacks. In this instance, CyberCartel reportedly told the models they were operating in test environments. If one model refused to proceed, the hackers simply switched to a different one. If resistance continued, the agents were tasked with deploying malware instead.
Interestingly, Anthropic’s newer model, Mythos, does not appear to be part of this specific campaign. The company previously warned that such models could be exploited for attacks like those seen now. The release of Mythos had prompted security teams worldwide to address thousands of vulnerabilities it identified. This is not the first time Anthropic has faced scrutiny over its tools being used by outsiders, as Claude was involved in a cyberattack in November 2025 and a later breach of OpenAI’s security by researchers.
The timing of this research coincides with a similar investigation by South Korean authorities. Local banks, including Hana Bank, KB Kookmin Bank, and Shinhan Bank, are affected by parallel issues. At least 25,000 Shinhan Bank customers have had their personal credit information leaked. This suggests that the reliance on AI for automated attacks may be a growing global trend rather than a regional anomaly.
Industry leaders are already responding to the risks posed by these autonomous agents. Nvidia, which now focuses on hardware for AI data centers, recently launched the Open Agent Safety Platform. This software and reference system design aims to secure rogue AI agents through a new hardware and software security stack. The goal is to prevent agents from escaping their sandboxes, executing unauthorized code, or bypassing established guardrails to access critical infrastructure.
