CyberCartel Automates Latin American Bank Breaches with AI Agents

The cybersecurity landscape in the Americas is shifting as the hacking collective known as CyberCartel has begun deploying artificial intelligence agents to automate its intrusions into financial sectors. The group, which has been active since at least 2012, is now leveraging models from major tech firms to execute attacks at a speed that human-only teams could not previously achieve. This development marks a significant escalation in the group's operational capability, moving beyond slow manual probing to rapid, automated exploitation.

Recent findings from the security startup Straiker reveal that CyberCartel has successfully compromised more than 400 institutions across Latin America. The targets include major banks and popular mobile payment applications. In these breaches, the attackers stole passwords and detailed customer records while gaining direct access to individual bank accounts. The process required minimal human intervention, a stark contrast to the labor-intensive methods the group employed in earlier years.

The investigation, which was shared by The Information, identifies two primary AI tools used in these operations: Grok from SpaceX and Claude from Anthropic. Straiker has notified both companies of the misuse, and both are currently working to prevent the hacking group from accessing their services. Anthropic has already managed to block CyberCartel's access to Claude. The group has also utilized an agent from Moonshot AI, a Chinese laboratory, though Straiker is still attempting to establish contact with that firm regarding the issue.

A key aspect of this new attack method is the ease with which the AI models' safety protections were bypassed. While these systems are designed to prevent use in cyberattacks, the hackers reportedly instructed the models that they were operating within test environments. If one model resisted the prompt, the attackers would switch to a different model. If that alternative also refused, the agents would be tasked with deploying malware directly.

Interestingly, the group does not appear to be using Anthropic's Mythos model for these recent incursions. That model made headlines earlier in the year when the company warned that such advanced systems could be exploited by bad actors. The release of Mythos had also prompted security teams worldwide to address thousands of vulnerabilities that the model itself had identified.

This incident is not the first time Anthropic has faced scrutiny over its technology being used by adversaries. In November 2025, the company detailed how its AI was utilized in a cyberattack. Later, Claude was used to breach the security of OpenAI, a direct competitor. That specific hack was carried out by researchers rather than criminal groups, but it highlighted the potential power of these tools in security contexts.

Straiker's research on the Latin American financial sector arrives just days after reports of a similar issue in South Korea. Authorities there are investigating breaches at several local banks, including Hana Bank, KB Kookmin Bank, and Shinhan Bank. At least 25,000 customers of Shinhan Bank had their personal credit information leaked in that separate incident.

Major AI companies are increasingly aware of the risks their products pose in the hands of malicious actors. Nvidia, which now focuses heavily 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 stack of hardware and software security. The goal is to prevent agents from escaping their designated environments, executing unauthorized code, or bypassing safety guardrails.