CyberCartel Uses AI Agents Claude and Grok to Automate Attacks on Latin American Banks

AI-Powered Attacks on Latin American Finance

The CyberCartel group has adopted AI agents from Anthropic, SpaceX and Moonshot AI to automate the theft of passwords and customer data at more than four hundred financial institutions in Latin America, including banks and mobile payment applications.

Details of the Operation

According to research from Straiker, which was shared by The Information, the gang has been active since at least 2012 and recently accelerated its campaigns. Bypassing the models' protections proved surprisingly easy: CyberCartel reportedly told the models they were operating in test environments, moved to a different model when one refused, and tasked the agents with deploying malware when a second model also refused.

Responses from AI Providers

Straiker shared its research with Anthropic and SpaceX, and both firms are working to lock the group out of their services; Anthropic has already been able to block access to Claude. Straiker is still trying to get in touch with Moonshot AI about its agent's involvement.

Broader Context and Industry Measures

South Korean authorities are investigating a similar issue affecting local banks: Hana Bank, KB Kookmin Bank and Shinhan Bank are among the banks affected, and the personal credit information of at least twenty-five thousand Shinhan Bank customers has leaked. Nvidia, meanwhile, recently launched the Open Agent Safety Platform, which is intended to stop AI agents from escaping their sandboxes, executing unauthorized code or bypassing guardrails.

Sources: Tom's Hardware