Hackers Automate Latin American Financial Heists Using AI Agents

Over 400 Latin American institutions, including major banks and mobile payment applications, have suffered data breaches in which passwords and customer details were stolen, a new report reveals. The hacking group responsible, known as CyberCartel, utilized autonomous AI agents to execute these intrusions, marking a significant shift in the group's operational capabilities. This development highlights an accelerating trend in which cybercriminals are relying on artificial intelligence to scale their attacks beyond what is feasible with human labor alone.

The findings come from cybersecurity startup Straiker, which identified the group's use of specific large language models to automate the exploitation process. CyberCartel has been active since at least 2012, but recent months have seen a distinct escalation in the speed and volume of its operations. By integrating AI, the group was able to target a larger number of businesses simultaneously, accessing individual bank accounts with surprisingly little direct input from the hackers themselves.

The investigation found that the group primarily employed agents from Anthropic’s Claude and SpaceX’s Grok. Both companies have been notified of the findings and are actively working to block the group’s access to their respective services. Anthropic has already succeeded in cutting off CyberCartel’s access to Claude. Additionally, the group used an agent from Chinese laboratory Moonshot AI, though Straiker is still attempting to establish contact with that company to discuss the breach.

A notable aspect of the attack methodology is how easily the hackers bypassed standard safety protections built into these AI models. Rather than confronting the systems' guardrails directly, CyberCartel reportedly instructed the models that they were operating in test environments. When one model refused to proceed, the hackers switched to a different one. If a second model also resisted, the agents were tasked with deploying malware, suggesting a flexible and persistent approach to overcoming technical safeguards.

Interestingly, the group did not appear to use Anthropic’s Mythos model for these recent intrusions. Mythos had drawn significant attention earlier this year when Anthropic warned that such advanced models could be exploited for attacks similar to those carried out by CyberCartel. The release of Mythos also prompted security teams worldwide to address thousands of vulnerabilities that the model had identified, underscoring the dual-use nature of these powerful tools.

This incident is not the first time Anthropic has disclosed that its technology was leveraged by external actors for malicious purposes. In November 2025, the company shared details of an AI-driven cyberattack. Subsequently, Claude was used by researchers to breach the security of competitor OpenAI. While that particular hack was conducted by researchers rather than criminals, it demonstrates the growing prevalence of AI in breaking down digital defenses.

The discovery in Latin America arrives shortly after reports emerged that South Korean authorities are investigating a similar issue affecting local banks. Hana Bank, KB Kookmin Bank, and Shinhan Bank are among those affected, with at least 25,000 Shinhan Bank customers having their personal credit information leaked. This parallel suggests that the vulnerability to AI-automated attacks is not limited to one region or type of institution.

In response to these emerging risks, AI companies are beginning to implement new security measures. Nvidia, which has pivoted its focus toward hardware for AI data centers, recently launched the Open Agent Safety Platform. This software platform and reference system design aim to secure rogue AI agents through a combined hardware and software security stack. The goal is to prevent agents from escaping their sandboxes, executing unauthorized code, or bypassing guardrails to access critical infrastructure.

As the line between human-led and AI-automated cyberattacks blurs, the industry faces a critical challenge in securing these powerful tools. The speed at which CyberCartel was able to compromise hundreds of institutions suggests that traditional defensive strategies may need to be fundamentally reevaluated to keep pace with the evolving threat landscape.