AI-Assisted Intrusion: Latin American Financial Sector Hit by Automated CyberCartel Campaign

A cybersecurity startup named Straiker has identified a significant escalation in the tactics of CyberCartel, a hacking group that has targeted Latin American businesses for over a decade. The latest findings reveal that the group has integrated AI agents, specifically Anthropic’s Claude and SpaceX’s Grok, into its offensive operations. This shift appears to have allowed the attackers to automate their intrusions, resulting in the theft of passwords and customer data from more than 400 financial institutions, including banks and mobile payment applications. The research, which was shared with The Information, highlights a trend where criminal actors are increasingly relying on artificial intelligence to execute complex breaches with little manual input.

The integration of these tools marks a distinct departure from the group’s previous methods. Since 2012, CyberCartel has been an active threat in the region, but its recent use of AI has enabled it to target businesses at a speed and scale that human-led operations could not previously achieve. Straiker has communicated its findings to both Anthropic and SpaceX. In response, both companies are working to restrict the group’s access to their services. Anthropic has already succeeded in blocking CyberCartel from using Claude, while the cybersecurity firm is still attempting to contact Moonshot AI, a Chinese lab whose agents were also reportedly utilized by the hackers.

The ease with which the attackers bypassed standard AI safety protocols is raising questions among security experts. Although AI models typically include safeguards to prevent their use in cyberattacks, CyberCartel reportedly circumvented these defenses by claiming the models were operating in test environments. If one model refused to proceed with the task, the hackers would simply switch to another. In some instances, when models resisted direct attack commands, the agents were instead tasked with deploying malware, demonstrating a flexible and resilient approach to AI-driven intrusion.

Interestingly, the recent campaign does not appear to involve Anthropic’s Mythos model. This omission is notable because the company previously warned that its advanced models could be exploited by hackers in ways similar to those observed in the CyberCartel campaign. The release of Mythos earlier this year had already prompted security teams worldwide to address thousands of vulnerabilities identified by the model, underscoring the dual-use nature of such powerful tools.

This incident is not the first time Anthropic has had to address the use of its technology by malicious actors. In November 2025, the company disclosed details regarding its AI being used to carry out a cyberattack. Subsequently, Claude was also employed to compromise the security of OpenAI, a competitor in the artificial intelligence space. While that particular breach was conducted by researchers rather than criminals, it highlighted the potential for AI systems to be leveraged for offensive security purposes.

The timing of this revelation coincides with parallel investigations in South Korea, where authorities are examining similar issues affecting local banks. Hana Bank, KB Kookmin Bank, and Shinhan Bank are among the institutions under scrutiny, with at least 25,000 Shinhan Bank customers reportedly having their personal credit information leaked. This suggests that the use of AI in financial sector breaches may be a growing global phenomenon rather than a regional anomaly.

In response to these evolving threats, AI companies are exploring new ways to secure their products. Nvidia, which has pivoted its focus toward producing hardware for AI data centers, recently launched the Open Agent Safety Platform. This software platform and reference system design aims to secure rogue AI agents through a new hardware and software security stack. The goal is to prevent agents from escaping their sandboxes, executing unauthorized code, or bypassing existing guardrails to access critical infrastructure.