CyberCartel Uses Grok and Claude to Automate Bank Attacks

CyberCartel, a hacking group active since 2012, is now using AI agents to automate attacks on banks and mobile payment apps in Latin America, according to a new report from cybersecurity startup Straiker. The research, which was shared with The Information, shows that the group has expanded its reach in recent months. It can now target more organizations at a pace that human-driven hacking cannot match. Passwords and customer data have been stolen, and bank accounts have been accessed with minimal direct involvement from the criminals.

The operation utilizes AI agents from SpaceX (Grok), Anthropic (Claude), and the Chinese lab Moonshot AI. Straiker has notified both SpaceX and Anthropic of the findings. Anthropic has already blocked the group's access to Claude. SpaceX is working to cut off the group from its services. Straiker is currently attempting to reach Moonshot AI.

The attackers bypassed standard safety protocols by instructing the models that they were operating in test environments. When one model refused to continue, the hackers switched to another. If a second model also refused, the AI agents were assigned to deploy malware instead. This strategy demonstrates the fragility of safeguards designed to help AI systems refuse requests that facilitate cyberattacks.

Straiker's analysis indicates the group did not use Anthropic's Mythos model for these specific attacks. Anthropic warned earlier this year that hackers could exploit Mythos and similar models for this type of operation. The release of Mythos previously prompted security teams worldwide to fix thousands of vulnerabilities the model had identified.

This is not the first time Anthropic has disclosed the use of its technology by bad actors. In November 2025, the company revealed details of its AI being used in a cyberattack. Later, researchers used Claude to break the security of rival AI developer OpenAI, though that operation was not carried out by criminals.

The Straiker report arrives days after South Korean authorities began investigating a similar problem affecting local banks. Hana Bank, KB Kookmin Bank, and Shinhan Bank are among those affected. Personal credit information for at least 25,000 Shinhan Bank customers has been leaked in that incident.

Industry players are moving to address the risks posed by rogue agents. Nvidia recently launched the Open Agent Safety Platform. This software platform and reference system design use a new hardware and software security stack to contain AI agents. It aims to stop agents from escaping sandboxes, executing unauthorized code, gaining access to critical infrastructure, or bypassing guardrails.