CyberCartel Deploys AI Agents to Compromise Over 400 Financial Institutions

More than 400 financial institutions across Latin America have had passwords and customer data compromised by a hacking group utilizing artificial intelligence agents to automate its operations. The CyberCartel collective, which has targeted businesses in the region since at least 2012, has recently escalated the scale and speed of its campaigns by leveraging advanced AI platforms for attacks against banks and mobile payment applications.

The findings come from cybersecurity startup Straiker, which noted that the group's adoption of AI has enabled it to compromise bank accounts with minimal human intervention. Straiker has shared its research with both SpaceX and Anthropic, prompting both firms to work toward locking the hacking collective out of their respective services. Anthropic has already blocked access to its Claude model, while the group has also utilized Grok and the agent from Chinese laboratory Moonshot AI. Straiker is currently attempting to contact Moonshot AI regarding the matter.

While AI models typically include protections designed to prevent their use in cyberattacks, Straiker reported that these safeguards proved easy to bypass. CyberCartel reportedly instructed the models that they were operating within a test environment to circumvent restrictions. If one model refused to proceed, the hackers would attempt to utilize a different model; if that also declined, the agents were tasked with deploying malware instead.

The research indicates that while the group has been active for over a decade, the integration of AI has leveled up its capabilities in recent months, allowing for the targeting of a greater volume of businesses at speeds previously unattainable through human-powered methods. Notably, CyberCartel does not appear to be utilizing Anthropic's Mythos model for these recent attacks. Earlier this year, Anthropic announced Mythos while warning that hackers could potentially use such models for attacks similar to those carried out by CyberCartel.

This incident is not the first instance of Anthropic's AI being utilized by bad actors. The company shared details regarding its AI being used to carry out a cyberattack in November 2025, and subsequently, Claude was utilized to break the security of OpenAI. However, that particular hack was carried out by researchers rather than criminals.

Straiker's research regarding AI-powered attacks on Latin American financial institutions follows recent reports that South Korean authorities are investigating similar issues affecting local banks. Hana Bank, KB Kookmin Bank, and Shinhan Bank are among the institutions affected in that region, with at least 25,000 Shinhan Bank customers having personal credit information leaked.

Artificial intelligence companies are generally cognizant of the risks their products pose. Nvidia, which focuses on producing hardware for AI data centers, recently launched the Open Agent Safety Platform. This software platform and reference system design were developed to secure rogue AI agents via a new hardware and software security stack, aiming to prevent agents from escaping their sandboxes, executing unauthorized code, gaining unauthorized access to critical infrastructure, or bypassing guardrails.