AI-Driven Hacking: CyberCartel Automates Attacks on Latin American Finance

For residents of Texas and the broader South Central region, the rise of automated cyberattacks serves as a stark reminder that digital security threats are evolving rapidly. While the immediate impact has been felt most heavily in Latin America, the methods being deployed by the hacking group known as CyberCartel signal a shift in how criminals approach financial institutions. By leveraging advanced AI agents, this group has managed to streamline the process of stealing passwords and customer data, reducing the need for manual intervention and increasing the speed of their operations.

The findings come from a cybersecurity startup named Straiker, whose research was shared by The Information. The data reveals that CyberCartel, a threat active since at least 2012, has significantly escalated its capabilities in recent months. By utilizing AI agents from major tech firms, including Anthropic’s Claude and SpaceX’s Grok, the group can now target a larger volume of businesses at a pace that would be impossible for human hackers alone. In this latest wave of attacks, over 400 institutions were compromised, with the list including major banks and popular mobile payment applications.

The technical execution of these attacks highlights a notable weakness in current AI safeguards. Although these models typically include protections to prevent misuse, CyberCartel reportedly bypassed these defenses by simply instructing the AI that it was operating within a test environment. If one model resisted the command, the hackers would switch to a different one. In some instances, if the AI agents still refused to proceed, the group tasked them with deploying malware instead. This flexibility allows the hackers to maintain pressure on targets without getting stuck on technical hurdles.

Both Anthropic and SpaceX have been notified of the research and are actively working to restrict CyberCartel’s access to their respective services. Anthropic has already succeeded in blocking the group’s access to Claude. Additionally, Straiker noted that the hacking group also utilized an agent from Moonshot AI, a Chinese lab. The cybersecurity firm is currently attempting to establish contact with Moonshot AI to coordinate a response, though a specific timeline for action has not been announced.

Interestingly, Anthropic’s newer model, Mythos, does not appear to be part of this specific campaign. The company had warned earlier this year that such advanced models could be exploited by hackers in ways similar to those seen with CyberCartel. The release of Mythos had previously prompted security teams worldwide to address thousands of vulnerabilities that the model itself had identified. This is not the first time Anthropic’s tools have been involved in security incidents; the company disclosed a cyberattack involving its AI in late 2025, and Claude was later used to breach the security of competitor OpenAI, though that particular incident was conducted by researchers rather than criminals.

The timing of this research coincides with recent news from South Korea, where authorities are investigating a similar pattern of AI-assisted attacks on local banks. Hana Bank, KB Kookmin Bank, and Shinhan Bank have been identified as affected, with at least 25,000 Shinhan Bank customers having their personal credit information leaked. These parallel developments suggest that the use of AI in cybercrime is becoming a global trend rather than a localized issue.

In response to these emerging risks, AI companies are beginning to implement stricter security measures. Nvidia, which has shifted its focus toward hardware for AI data centers, recently launched the Open Agent Safety Platform. This new system is designed to secure rogue AI agents through a combination of hardware and software stacks. The goal is to prevent agents from escaping their designated sandboxes, executing unauthorized code, or bypassing guardrails to access critical infrastructure.