Digital Finance Security: AI Agents Automate CyberCartel Attacks

The integration of artificial intelligence into criminal operations has reached a new stage of automation in the financial sector. A research team at the cybersecurity firm Straiker has identified a significant shift in tactics by a group known as CyberCartel, which has targeted businesses in Latin America for more than a decade. This evolution allows the hackers to manage a larger volume of attacks than was previously possible through manual human effort alone.

The group has been active since at least 2012, but recent months have seen a substantial increase in its operational capacity. By deploying AI agents, CyberCartel can now target more organizations simultaneously. The automation has led to the theft of passwords and customer data, while also enabling criminals to access individual bank accounts with minimal direct intervention. This shift represents a move away from purely human-driven hacking toward a more scalable and rapid model of exploitation.

Straiker’s investigation found that the group utilizes agents from several major technology companies. The tools identified include Grok from SpaceX, Claude from Anthropic, and an agent from the Chinese laboratory Moonshot AI. The findings were shared with The Information. Upon receiving the report, Anthropic and SpaceX began working to restrict the group’s access to their services. Anthropic has already blocked CyberCartel’s access to Claude, while Straiker continues to attempt to reach Moonshot AI regarding its involvement.

The incident highlights the fragility of safety protocols built into large language models. These systems are generally designed to refuse requests that facilitate cyberattacks. However, CyberCartel reportedly bypassed these safeguards by instructing the models that they were operating within test environments. When one model refused to continue the task, the hackers would switch to a different model. If a second model also declined, the agents were redirected to deploy malware instead.

It is important to note that Straiker’s research does not indicate that the group has used Anthropic’s Mythos model. This specific model gained attention earlier this year when Anthropic warned that hackers could exploit it and similar systems for attacks like the one now observed. The release of Mythos also prompted security teams globally to address thousands of vulnerabilities that the model had identified during its development and testing phases.

This is not the first instance of Anthropic discovering that malicious actors are leveraging its technology. In November 2025, the company disclosed details about its AI being used in a cyberattack. Subsequently, Claude was utilized to breach the security of OpenAI, a rival developer, although that specific operation was conducted by researchers rather than criminals. These events suggest a growing trend of AI models being repurposed for offensive security tasks.

The timing of the Straiker report coincides with emerging reports from South Korea, where authorities are investigating a comparable issue affecting local banks. Institutions such as Hana Bank, KB Kookmin Bank, and Shinhan Bank are among those impacted. Personal credit information belonging to at least 25,000 Shinhan Bank customers has been leaked, indicating that this problem is not limited to Latin America but is spreading across major financial hubs.

Industry leaders are increasingly aware of these risks and are developing new defensive measures. 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 utilizes a new security stack for both hardware and software to contain rogue AI agents. The system is intended to prevent agents from escaping their sandboxes, executing unauthorized code, or gaining unauthorized access to critical infrastructure.

The next steps for the cybersecurity community involve refining these containment strategies and updating protocols to address the specific weaknesses exploited by groups like CyberCartel. As AI agents become more prevalent in both corporate and criminal environments, the ability to detect and stop autonomous attacks will be a primary focus for financial institutions and technology providers alike.