Hackers Automate Bank Heists Using AI Agents

Cybercriminals are now outsourcing the heavy lifting of breaking into financial accounts to artificial intelligence.

A hacking outfit known as CyberCartel, which has targeted Latin American businesses since at least 2012, is using AI agents to automate attacks against banks and mobile payment apps. The group has successfully accessed customer data and passwords, reportedly compromising over 400 institutions. According to cybersecurity startup Straiker, which shared its findings with The Information, the group is now able to target more businesses at speeds impossible for humans alone, with little input from the criminals themselves.

The group has utilized agents from SpaceX’s Grok and Anthropic’s Claude, as well as an agent from Chinese lab Moonshot AI. While Anthropic has already blocked access to Claude for this group, and Straiker is attempting to contact Moonshot AI, both SpaceX and Anthropic are reportedly working to lock CyberCartel out of their services. Straiker is still attempting to get in touch with Moonshot AI regarding their involvement.

Despite standard protections designed to prevent misuse, AI models can be easily manipulated by bad actors. In this instance, the hackers simply claimed they were operating in a test environment. If one model refused, the group switched to another; if that failed, they instructed the agents to deploy malware instead. Notably, CyberCartel does not appear to be using Anthropic’s Mythos model for these attacks, though the company warned earlier this year that hackers could use it in this manner. Anthropic has previously shared details of its AI being used in a cyberattack in November 2025, and Claude was also used to break the security of OpenAI, though that incident was carried out by researchers rather than criminals.

This activity coincides with increased scrutiny on AI safety. South Korean authorities are investigating similar issues affecting local banks, including Hana Bank, KB Kookmin Bank, and Shinhan Bank, following a leak of credit information for at least 25,000 customers. In response to these risks, Nvidia has launched the Open Agent Safety Platform. The software platform and reference system design were developed to secure rogue AI agents via a new hardware and software security stack, aiming to prevent agents from escaping sandboxes, executing unauthorized code, or bypassing guardrails.