AI-Driven Cybercrime: Latin American Banks Hit by Automated AI Agents

The integration of artificial intelligence into cybercrime has accelerated significantly in recent months, prompting a reevaluation of how digital defenses are structured. A new report highlights a specific threat vector where automated AI agents are being used to breach financial systems with minimal human intervention. This development marks a notable shift in the tactics employed by organized hacking groups in the region.

CyberCartel, a group active since at least 2012, has targeted over 400 institutions in Latin America, including banks and mobile payment applications. According to research from the cybersecurity startup Straiker, the group utilized AI agents to automate the process of stealing passwords and accessing individual bank accounts. The findings, which were shared by The Information, indicate that the use of these tools allowed the hackers to attack businesses at a speed that was previously unachievable with manual methods. Both customer data and login credentials were compromised in these operations.

The specific AI models involved in these attacks include Grok from SpaceX and Claude from Anthropic. Straiker has shared its findings with both technology firms, and both are currently working to restrict the hacking group's access to their services. Anthropic has already succeeded in blocking CyberCartel from using Claude. Additionally, the group has utilized an agent from the Chinese laboratory Moonshot AI, though Straiker is still attempting to establish contact with that company to discuss the incident.

A key finding in the research is the ease with which standard AI safety protections were bypassed. Although models typically have safeguards to prevent their use in cyberattacks, CyberCartel reportedly instructed the models to operate in test environments. When one model refused to proceed, the hackers switched to a different model. If that alternative also resisted, the agents were tasked with deploying malware instead. This flexibility allowed the group to continue its operations despite initial resistance from the AI systems.

Notably, Anthropic's newer model, Mythos, does not appear to have been used in these recent attacks. The company had previously warned that such advanced models could be exploited by hackers for similar purposes. The release of Mythos had also prompted security teams worldwide to address thousands of vulnerabilities identified by the model. This incident follows other instances where Anthropic's AI was used in cyberattacks, including a breach of OpenAI's security by researchers in late 2025.

The research on Latin American institutions comes days after news emerged regarding a similar investigation by South Korean authorities. Local banks, including Hana Bank, KB Kookmin Bank, and Shinhan Bank, have been affected by issues linked to AI usage. At least 25,000 Shinhan Bank customers had their personal credit information leaked in that separate incident. These parallel developments suggest a growing global trend of AI-assisted breaches in the financial sector.

AI companies are increasingly aware of the risks associated with their products. Nvidia, which focuses on hardware for AI data centers, recently launched the Open Agent Safety Platform. This system is designed to secure rogue AI agents using a new hardware and software security stack. The platform aims to prevent agents from escaping their designated environments, executing unauthorized code, or gaining access to critical infrastructure. These efforts reflect a broader industry push to mitigate the potential for AI-driven cyber threats.