CyberCartel Automates Attacks With AI Agents

A long-running cybercrime operation targeting Latin American financial systems has adopted artificial intelligence to automate its intrusions. Research from Straiker, shared via The Information, indicates the group known as CyberCartel is leveraging AI agents to compromise banks and mobile payment applications. Passwords and customer data have been compromised, and bank accounts have been accessed with minimal manual effort from the perpetrators.

CyberCartel has been an active threat against Latin American businesses since at least 2012. The group's recent adoption of AI technology has elevated its capabilities, allowing it to target more institutions at a pace unattainable through traditional methods. Straiker has notified SpaceX and Anthropic of its findings, and both companies are taking action to prevent the group from accessing their platforms. Anthropic has already restricted access to Claude. The hacking group has also utilized an agent from Moonshot AI, though Straiker has not yet contacted that company.

The hackers bypassed standard model protections by asserting they were operating in test environments. If one model declined the request, the group switched to a different model. Should the second model also refuse, the agents were instructed to deploy malware instead. Notably, the group is not utilizing Anthropic's Mythos model for these attacks, despite Anthropic's earlier warnings that such tools could be used for malicious purposes.

The research comes days after South Korean authorities launched an investigation into similar incidents involving local banks. Hana Bank, KB Kookmin Bank, and Shinhan Bank were named as affected institutions, with over 25,000 Shinhan Bank customers having their personal credit information exposed. Anthropic previously disclosed in November 2025 that its AI was used to carry out a cyberattack. Subsequently, Claude was used to break OpenAI's security by researchers.

AI companies are actively developing safeguards for their products. Nvidia recently launched the Open Agent Safety Platform to secure rogue AI agents through a new hardware and software security stack. The system aims to prevent agents from escaping sandboxes, executing unauthorized code, gaining unauthorized access to critical infrastructure, or bypassing guardrails.