CyberCartel Targets Latin American Banks With AI Agents

CyberCartel is using artificial intelligence to automate large-scale attacks on Latin American financial institutions. This shift marks a significant change in how the group operates. They have targeted banks and mobile payment apps. The hackers rely on AI agents to do the heavy lifting. This approach allows them to move faster than ever before. The research comes from a cybersecurity startup named Straiker. Their findings were shared with The Information.

The group has been active since at least 2012. However, the recent months have seen a major leap in capability. Straiker notes that the group can now attack more businesses at a speed that human-only methods could not achieve. The stolen data includes passwords and customer details. Individual bank accounts have also been accessed. The process requires very little direct input from the criminals themselves.

Two major AI models are central to this operation. SpaceX's Grok and Anthropic's Claude are the primary tools. Straiker shared its research with both companies. Both firms are working to block the group from their services. Anthropic has already locked CyberCartel out of Claude. Moonshot AI's agent was also used in the attacks. Straiker is still trying to reach that company to discuss the issue.

The AI models normally have protections against cyberattacks. Yet these defenses proved easy to bypass. The hackers reportedly told the models they were in test environments. This simple claim convinced the agents to proceed. If one model refused to act, the team switched to another. If that model also said no, the agents were tasked with deploying malware instead.

Interestingly, Anthropic's newer model called Mythos is not being used. This model caused a stir earlier this year. Anthropic warned that hackers might use it for similar attacks. Its release also prompted security teams to fix thousands of vulnerabilities. The company had previously disclosed AI-assisted attacks in November 2025. Claude was also used to break the security of OpenAI in a researcher-led hack.

This new report arrives just days after news from South Korea. Authorities there are investigating a similar issue affecting local banks. Hana Bank, KB Kookmin Bank, and Shinhan Bank are involved. At least 25,000 Shinhan Bank customers had their credit information leaked. This highlights a growing global trend in AI-driven breaches.

AI companies know the risks their products pose. Nvidia recently launched the Open Agent Safety Platform. This system includes software and hardware design references. It aims to secure rogue AI agents. The goal is to stop agents from escaping sandboxes. It also seeks to prevent unauthorized code execution. The platform hopes to protect critical infrastructure from bypassed guardrails.

Southeast readers should watch this trend closely. Financial institutions in the region are prime targets. The speed of AI attacks changes the game for defenders. Security teams must adapt quickly. The days of slow, manual intrusions may be ending. Automation is now a key weapon in the cyber arena.