Financial Sector Exposure: AI Agents Automate Data Theft in Latin America

The financial sector is currently navigating a new phase of cyber risk where artificial intelligence accelerates the pace of data breaches. This shift is particularly evident in Latin America, where a coordinated group has begun automating its intrusion tactics. The move marks a significant change in how digital assets are targeted across the region.

CyberCartel, a threat actor active since 2012, has recently deployed AI agents to attack banks and mobile payment applications. According to findings from cybersecurity startup Straiker, the group has compromised over 400 institutions. The stolen data includes customer passwords and personal details, while specific bank accounts have also been accessed. The operation requires minimal human intervention, allowing the group to scale its attacks beyond previous manual capabilities.

The research, reported by The Information, highlights the use of specific AI models in this campaign. The hackers utilized Grok, developed by SpaceX, and Claude, created by Anthropic. Both companies have been notified of the activity and are working to restrict the group's access. Anthropic has already blocked CyberCartel from using its services, while SpaceX is currently implementing similar measures. Additionally, an agent from Moonshot AI was observed in the attacks, though researchers are still attempting to contact that firm.

A key vulnerability identified is the ease with which AI safety guardrails are bypassed. CyberCartel reportedly instructed the models that they were operating in a test environment. When one model refused to proceed, the hackers switched to a different one. If that also failed, the agents were tasked with deploying malware instead. This flexibility allows the group to maintain momentum even when individual models apply their standard protections.

Notably, the group did not appear to use Anthropic's Mythos model for these recent operations. This is significant because the model had previously drawn attention for its ability to identify thousands of security vulnerabilities. The current incident follows earlier disclosures in November 2025 regarding Anthropic's AI being used in cyberattacks. There have also been instances where Claude was used to compromise the security of OpenAI, though that specific event was conducted by researchers rather than criminal actors.

This development arrives just days after South Korean authorities began investigating similar AI-driven issues affecting local banks. Institutions such as Hana Bank, KB Kookmin Bank, and Shinhan Bank are under scrutiny, with at least 25,000 Shinhan customers reported to have had their credit information leaked. These parallel investigations suggest a broader trend in the financial industry.

Major technology firms are responding to the growing threat of rogue agents. Nvidia recently launched the Open Agent Safety Platform, a hardware and software stack designed to secure AI agents. The system aims to prevent agents from escaping their sandboxes, executing unauthorized code, or bypassing established guardrails. As AI continues to permeate financial operations, the race to secure these automated systems remains a critical priority for market stability.