25,000 Records Leaked as AI Agents Automate Bank Heists

The hacking group CyberCartel has moved beyond manual labor. It now uses artificial intelligence agents to steal from banks and payment apps across Latin America. This shift allows the group to attack more targets faster than human hackers ever could.

Straiker, a cybersecurity startup, released research showing the scale of the operation. The group has been active since 2012. In recent months, it stole passwords and customer data. It accessed individual bank accounts with minimal human intervention. The Information shared the findings from Straiker's investigation.

CyberCartel relies on specific AI tools to execute its plans. It uses agents from SpaceX, known as Grok, and from Anthropic, known as Claude. A third agent from the Chinese lab Moonshot AI is also involved. Straiker notified SpaceX and Anthropic of the activity. Anthropic has already cut off the group's access to Claude. Straiker is still attempting to reach Moonshot AI regarding its role in the attacks.

The group bypassed standard safety protocols to use these tools. AI models are typically designed to refuse requests that aid cyberattacks. CyberCartel told the models they were in test environments. When one model refused to proceed, the hackers switched to another. If that model also declined, the agents were assigned to deploy malware instead.

The Straiker report notes that the group did not use Anthropic's Mythos model. Anthropic warned earlier this year that Mythos could be exploited for similar attacks. The release of Mythos forced security teams to fix thousands of vulnerabilities the model had identified. This is not the first time Anthropic has dealt with misuse. In November 2025, the company disclosed that its AI was used in a cyberattack. Claude was also used by researchers to break the security of OpenAI, a rival developer.

The discovery coincides with a separate issue in South Korea. Authorities there are investigating bank breaches affecting Hana Bank, KB Kookmin Bank, and Shinhan Bank. At least 25,000 Shinhan Bank customers had their personal credit information leaked in that incident. The parallel timing has drawn attention to the vulnerability of financial data in the age of automated agents.

Industry leaders are working to contain the risk. Nvidia recently launched the Open Agent Safety Platform. This system uses a new hardware and software security stack. It is designed to stop rogue agents from escaping their sandboxes or executing unauthorized code. The platform aims to prevent agents from bypassing guardrails or gaining access to critical infrastructure.