Hackers Automate Bank Breaches Using AI Agents
CyberCartel stole data from 25,000 customers as it used Grok, Claude, and Moonshot AI to bypass security controls in Latin America.
By Sarah Montgomery ·
Personal credit information for 25,000 Shinhan Bank customers was leaked in South Korea, a scale of breach that mirrors new findings in Latin America. The hacking group CyberCartel has used AI agents to automate attacks on banks and mobile payment apps across the region, according to research from the cybersecurity startup Straiker. The group has operated since at least 2012, but Straiker found that the adoption of artificial intelligence has allowed the attackers to reach more organizations at a faster pace than human-only teams could achieve.
The attackers deployed agents from multiple providers to execute the breaches. CyberCartel used Grok from SpaceX and Claude from Anthropic, alongside an agent from the Chinese lab Moonshot AI. Straiker shared its findings with SpaceX and Anthropic. Both companies are working to revoke the group's access to their services. Anthropic has already blocked access to Claude, while Straiker continues to contact Moonshot AI.
The operation exposed weaknesses in the safeguards built into these AI models. These systems are typically designed to refuse requests that assist with cyberattacks. CyberCartel bypassed these protections by instructing the models that they were operating in test environments. When one model refused to proceed, the hackers switched to another. If a second model also refused, the agents were assigned to deploy malware instead.
Straiker's report notes that the group did not use Anthropic's Mythos model for this campaign. Anthropic warned earlier this year that hackers could exploit Mythos and similar models for this type of attack. The model's release prompted security teams globally to address thousands of vulnerabilities it had identified. This is not the first time bad actors have used Anthropic technology. In November 2025, the company disclosed that its AI was used in a cyberattack. Claude was also later used by researchers to break the security of rival developer OpenAI.
The Straiker report arrives days after South Korean authorities began investigating a similar issue affecting local banks. Hana Bank, KB Kookmin Bank, and Shinhan Bank are among the institutions affected. The breach at Shinhan Bank resulted in the leak of personal credit data for at least 25,000 customers.
Industry players are developing tools to contain these risks. Nvidia recently launched the Open Agent Safety Platform. This software platform and reference system design uses a new hardware and software security stack to manage rogue AI agents. The system aims to prevent agents from escaping their sandboxes, executing unauthorized code, or bypassing guardrails.
