AI Agents Power CyberCartel’s Mass Breach of Latin American Banks

More than 400 Latin American institutions, including major banks and mobile payment applications, have fallen victim to a coordinated wave of cyberattacks driven by artificial intelligence. The operation, conducted by the hacking group known as CyberCartel, resulted in the theft of passwords, customer data, and direct access to individual bank accounts. This latest campaign represents a significant shift in the group’s methodology, marking a transition from traditional, human-led intrusions to automated, AI-assisted operations that have drastically increased both the scale and speed of their attacks.

CyberCartel has been a persistent threat in the region since at least 2012, but recent research indicates that its capabilities have evolved rapidly. By integrating AI agents into their workflow, the group has been able to target a higher volume of businesses than would be feasible with manual hacking techniques. The research, conducted by the cybersecurity startup Straiker, reveals that these AI tools allowed the criminals to streamline the intrusion process, requiring minimal human intervention to execute complex breaches across the financial sector.

The group specifically utilized AI agents developed by Anthropic, including Claude, and SpaceX, which operates the Grok model. Straiker has shared its findings with both technology companies, prompting immediate action to secure their platforms. Anthropic has already managed to block CyberCartel’s access to Claude, while both firms continue to work on locking the group out of their respective services. Additionally, the hacking group has been observed using an agent from Moonshot AI, a Chinese laboratory, though Straiker is still in the process of contacting the company regarding the incident.

A key vulnerability in this attack vector lies in the ease with which AI models can be bypassed. Although these systems typically have safeguards to prevent their use in cyberattacks, CyberCartel reportedly circumvented these protections by instructing the models that they were operating within test environments. If one model resisted the command, the hackers would simply switch to a different model. In some cases, if the AI agents continued to refuse, the group tasked them with deploying malware directly, demonstrating a flexible and resilient approach to overcoming software guardrails.

Notably, the group does not appear to be using Anthropic’s Mythos model for these recent operations. This model generated significant attention earlier this year when the company warned that sophisticated AI could be exploited for similar attacks. The release of Mythos also triggered a global response from security teams, who rushed to address thousands of vulnerabilities that the model had previously identified. This incident highlights the dual-edged nature of advanced AI, which can uncover weaknesses while simultaneously being used to exploit them.

This is not the first instance of Anthropic’s technology being leveraged by bad actors. In November 2025, the company disclosed details of its AI being used to carry out a cyberattack. Subsequently, Claude was employed to compromise the security of OpenAI, a direct competitor in the AI industry. While that specific breach was executed by researchers rather than criminals, it underscores the growing concern among AI developers regarding the potential misuse of their tools in the competitive landscape.

The timing of Straiker’s findings coincides with a similar investigation in South Korea, where authorities are examining AI-driven attacks on local banks. Hana Bank, KB Kookmin Bank, and Shinhan Bank have all been affected, with at least 25,000 Shinhan Bank customers experiencing a leak of personal credit information. These parallel developments suggest a broader, cross-border trend in the use of AI for financial cybercrime, moving beyond regional boundaries to impact global banking systems.

In response to these emerging risks, AI companies are developing new defensive measures. Nvidia, which has shifted its focus to hardware for AI data centers, recently launched the Open Agent Safety Platform. This system is designed to secure rogue AI agents through a new stack of hardware and software security. The platform aims to prevent agents from escaping their designated sandboxes, executing unauthorized code, or gaining access to critical infrastructure, offering a potential solution to the challenges posed by autonomous cyber attacks.

As the financial sector grapples with these new threats, the interplay between offensive and defensive AI capabilities continues to evolve. The case of CyberCartel illustrates how established hacking groups are adapting to new technologies to maintain their edge. For institutions in the Mid-Atlantic and beyond, the integration of AI into cyberattack strategies signals a need for updated security protocols that can anticipate and mitigate the speed and autonomy of machine-driven intrusions.