AI Agents: CyberCartel's New Weapon Against Financial Institutions

Customers of banks and mobile payment apps across Latin America could be at risk from a hacking group that is using artificial intelligence to speed up its attacks. A recent report details how the group, known as CyberCartel, is employing AI agents to target financial institutions, leading to the theft of passwords and customer data from over 400 organizations. The research also indicates that individual bank accounts have been accessed with minimal effort from the criminals.

CyberCartel has been a persistent threat to businesses in the region since at least 2012, but its recent adoption of AI has significantly upgraded its capabilities. According to findings from cybersecurity startup Straiker, the group can now launch attacks against a much larger number of targets at a rate that was not achievable through manual efforts. The research, which was shared by The Information, highlights the increasing automation of cybercrime.

The group's toolkit includes AI agents from Anthropic's Claude and SpaceX's Grok. Straiker has shared its findings with both companies, which are now working to block CyberCartel's access to their platforms. Anthropic has already succeeded in blocking access to its Claude model. The hacking group has also utilized an agent from Chinese developer Moonshot AI, though Straiker is still attempting to contact that company.

While AI models typically include safeguards to prevent misuse for malicious purposes like hacking, these protections have proven straightforward to circumvent. CyberCartel reportedly bypassed the restrictions by simply claiming the models were being used in a test environment. If one model refused to comply with the request, the group would switch to another. If that model also refused, the AI agents would then be tasked with deploying malware instead.

This incident is part of a broader pattern of AI models being used for malicious activities. Anthropic previously disclosed in November 2025 that its technology was used to conduct a cyberattack. In another instance, its Claude model was used to break the security of competitor OpenAI, though that breach was carried out by researchers rather than criminals. The Straiker report on Latin American financial institutions follows news of a similar investigation in South Korea, where the personal credit information of at least 25,000 Shinhan Bank customers was leaked.

In response to these emerging threats, technology companies are developing new security measures. Nvidia recently launched the Open Agent Safety Platform, a software and hardware security system designed to secure AI agents. The platform aims to prevent agents from escaping their designated environments, executing unauthorized code, or gaining unauthorized access to critical infrastructure.