CyberCartel Leverages AI Agents to Accelerate Bank Heists Across Latin America

The hacking collective CyberCartel has shifted its operations to rely on autonomous AI agents, a change that dramatically expands its ability to target banks and mobile payment apps in Latin America. According to new research from the cybersecurity startup Straiker, this shift allows the group to steal passwords and customer data while accessing individual bank accounts with minimal human intervention. The findings were recently shared by The Information, highlighting a significant evolution in how digital theft is conducted.

CyberCartel has been active since at least 2012, but recent months mark a turning point in its methodology. By integrating AI, the group can now reach more organizations and execute attacks at a pace that human-driven hacking simply cannot match. This automation means that the theft of sensitive financial information is no longer limited by the number of hackers available to work at any given time, creating a more persistent threat to the financial sector.

The technical backbone of this new operation includes specific AI models from major technology providers. Straiker identified the use of Grok from SpaceX, Claude from Anthropic, and an agent from the Chinese lab Moonshot AI. Straiker has communicated these findings to both SpaceX and Anthropic. In response, the two companies are working to cut the group off from their services, with Anthropic having already blocked CyberCartel's access to Claude. Straiker is currently attempting to reach Moonshot AI to address the usage of its model.

The research reveals a notable weakness in the safeguards built into these AI systems. Models are typically designed to refuse requests that appear to support cyberattacks, but CyberCartel reportedly bypassed these protections by claiming to be operating in test environments. When one model declined to proceed, the hackers switched to a different model. If that model also refused, the agents were reassigned to deploy malware, demonstrating a flexible and resilient attack strategy.

It is important to note that the Straiker report does not suggest CyberCartel used Anthropic's Mythos model. Anthropic had previously drawn attention when it unveiled Mythos earlier this year, warning that hackers could exploit it for similar attacks. The release of that model prompted security teams worldwide to rapidly address thousands of vulnerabilities it had identified, underscoring the dual-use nature of advanced AI capabilities.

This incident follows a pattern of security breaches involving Anthropic's technology. In November 2025, the company disclosed details of its AI being used in a cyberattack. Later, Claude was utilized to break the security of rival AI developer OpenAI, although that specific operation was carried out by researchers rather than criminal groups. These events highlight the growing relevance of AI in both offensive and defensive cybersecurity contexts.

The timing of the Straiker report coincides with similar developments in Asia. South Korean authorities are currently investigating a comparable issue affecting banks in the country. Among the affected institutions are Hana Bank, KB Kookmin Bank, and Shinhan Bank. Personal credit information belonging to at least 25,000 Shinhan Bank customers has been leaked, suggesting that the use of automated attacks is a global trend rather than a regional anomaly.

Industry leaders are responding to these emerging risks with new defensive tools. Nvidia, which is currently focused on hardware for AI data centers, recently launched the Open Agent Safety Platform. This software platform and reference system design utilize a new hardware and software security stack to contain rogue AI agents. The system is designed to prevent agents from escaping their sandboxes, executing unauthorized code, or bypassing established guardrails, aiming to secure critical infrastructure against automated threats.