CyberCartel Deploys AI Agents to Target Financial Systems

A hacking organization that has targeted Latin American businesses for over a decade is now automating its operations with commercial AI tools. CyberCartel, active since at least 2012, has recently begun using agents from SpaceX's Grok and Anthropic's Claude to infiltrate financial targets. The group is targeting banks and mobile payment apps, with the goal of stealing passwords and accessing customer data. This activity marks a significant escalation in the tactics used by the organization, which has historically relied on human-driven methods for its operations.

Research conducted by the cybersecurity startup Straiker, shared by The Information, quantifies the scale of the threat. The findings indicate that passwords and customer data have been stolen from more than 400 institutions, including banks and payment applications. The report also states that individuals' bank accounts have been accessed with minimal input from human criminals. This represents a new speed for the group, allowing it to target a greater number of businesses than was previously possible.

The involvement of major AI providers has prompted immediate responses. Straiker shared its findings with both SpaceX and Anthropic. Both companies are now working to prevent CyberCartel from accessing their services. Anthropic has already blocked Claude's access for the group. The hacking collective also used an agent from the Chinese lab Moonshot AI, though Straiker was still attempting to contact that company as of October 23.

The method CyberCartel used to bypass safety measures was straightforward. The models are generally designed to refuse requests to perform cyberattacks. However, the hackers reportedly convinced the models they were operating in a test environment. If one model still refused, they would switch to another. If that model also declined, they would task the AI with deploying malware instead. This tactic highlights a vulnerability in current safeguards. The group does not appear to be using Anthropic's Mythos model for these attacks.

This incident is part of a broader pattern of criminal misuse of AI tools. In November 2025, Anthropic reported its AI being used in a cyberattack. More recently, Claude was used by researchers to break the security of OpenAI. Security concerns are also escalating globally. South Korean authorities are investigating similar attacks affecting local banks, including Hana Bank, KB Kookmin Bank, and Shinhan Bank, with at least 25,000 Shinhan Bank customers having their credit information leaked.

The technology industry is responding to these threats. Nvidia, which produces 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, accessing critical infrastructure, or bypassing guardrails.