CyberCartel Exploits AI Agents to Automate Attacks on Latin American Financial Institutions

The personal credit information of at least 25,000 customers from Shinhan Bank in South Korea has been leaked, a figure that highlights the growing scale of threats facing the financial sector. This breach is part of a broader trend where cybercriminals are leveraging advanced technologies to expand their reach and efficiency in the global market.

Research from the cybersecurity startup Straiker reveals that a hacking group known as CyberCartel has significantly increased its operational capacity by deploying artificial intelligence agents. Active since at least 2012, the group has traditionally targeted businesses in Latin America. In recent months, the integration of AI has allowed the organization to pursue more entities at a pace that human-driven methods could not achieve. The automation has led to the theft of passwords and customer data, with individual bank accounts being accessed with minimal direct involvement from the criminals.

The operation utilizes agents from several prominent technology firms, including Grok from SpaceX and Claude from Anthropic. An agent from the Chinese lab Moonshot AI has also been employed in the campaign. Straiker shared its findings with both SpaceX and Anthropic. Anthropic has already blocked the group's access to its Claude model, while the two companies continue to work to sever the connection. Straiker is currently attempting to reach Moonshot AI regarding the usage of its tools.

The incident underscores the potential fragility of safeguards within AI models. These systems are generally designed to refuse requests that facilitate cyberattacks. However, CyberCartel reportedly bypassed these defenses by instructing the models that they were operating in test environments. When one model declined to proceed, the hackers switched to another. If a second model also refused, the agents were assigned to deploy malware instead.

The Straiker report clarifies that the group has not used Anthropic's Mythos model. This is notable because Anthropic previously warned that such models could be exploited for this type of attack. The release of Mythos had prompted security teams worldwide to address thousands of vulnerabilities identified by the system. This is not the first time Anthropic has disclosed that its technology was used in a cyberattack, as the company revealed similar details in November 2025.

The findings from Straiker arrive shortly after South Korean authorities began investigating a comparable issue affecting local banks. Hana Bank and KB Kookmin Bank are also among those affected by these developments. The situation has drawn attention to the need for robust security measures in the rapidly evolving landscape of digital finance.

Industry leaders are aware of these emerging risks. 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 employs a new hardware and software security stack 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.