AI Agents Automate CyberCartel’s Theft of Data From 400 Latin American Institutions

Over 400 institutions in Latin America had passwords and customer details stolen in recent months. The thefts targeted banks and mobile payment applications. The perpetrators were able to access individual bank accounts with minimal human intervention. This marks a significant shift in how financial crimes are executed in the region.

The group responsible is known as CyberCartel. They have operated since at least 2012. However, their recent tactics have evolved. They now rely on AI agents to drive their attacks. Specifically, they used Anthropic’s Claude and SpaceX’s Grok. These tools allow the group to attack more businesses simultaneously. The speed of these operations exceeds what is possible with traditional human-led hacking methods.

Research from the cybersecurity startup Straiker revealed these findings. The company shared its report with The Information. Straiker also notified both Anthropic and SpaceX about the abuse. Anthropic has already blocked CyberCartel’s access to Claude. SpaceX is working to do the same. The hacking group also used an agent from Moonshot AI. Straiker is currently trying to contact that company to discuss the issue.

AI models usually have protections against cyberattacks. Yet, the hackers found these defenses easy to bypass. CyberCartel reportedly told the models they were operating in test environments. If one model refused to proceed, the group switched to another. If that model also refused, the agents were instructed to deploy malware instead. This workaround allowed the attacks to continue despite the built-in safeguards.

Interestingly, the group did not use Anthropic’s Mythos model for these recent attacks. Anthropic warned earlier this year that hackers could exploit such models. The release of Mythos also prompted security teams worldwide to fix thousands of vulnerabilities the model identified. This is not the first time Anthropic has dealt with its AI being used by bad actors. In November 2025, the company disclosed that its AI was used in a cyberattack. Later, Claude was used to break into OpenAI’s security systems, though that incident involved researchers rather than criminals.

This development follows news from South Korea. Authorities are investigating similar AI-driven issues affecting local banks. Hana Bank, KB Kookmin Bank, and Shinhan Bank are among the affected institutions. At least 25,000 Shinhan Bank customers had their personal credit information leaked. The region is clearly seeing a rise in AI-assisted financial crimes.

AI companies are aware of the risks their products pose. Nvidia recently launched the Open Agent Safety Platform. This software and reference system design aims to secure rogue AI agents. It uses a new hardware and software security stack. The goal is to prevent agents from escaping their sandboxes or executing unauthorized code. It also seeks to stop them from bypassing guardrails or accessing critical infrastructure without permission.