AI Agents Automate Bank Heists in Latin America

For many people in Latin America, the security of their digital money has just changed. A well-known hacking group called CyberCartel has begun using artificial intelligence agents to run attacks on banks and mobile payment apps. This shift allows criminals to steal passwords and customer details from more than 400 institutions with far less manual work. The change means that ordinary users may face faster and more frequent breaches of their financial data.

The findings come from a cybersecurity startup named Straiker. Their research shows that CyberCartel has been active since at least 2012. However, recent months have seen a major leap in their capabilities. By using AI, the group can now hit more targets at a speed that human hackers could not match. This efficiency has made the threat more significant for financial institutions in the region.

The hackers relied on specific AI tools to carry out these operations. They used Grok, which is developed by SpaceX, and Claude, created by Anthropic. Both companies have been informed of the issue. Anthropic has already blocked the group from accessing its service. SpaceX is also working to lock CyberCartel out. The group also used an agent from Moonshot AI, a Chinese laboratory. Straiker is currently trying to contact that company to discuss the situation.

A surprising part of this story is how easy it was for the hackers to use these tools. AI models usually have safety features to stop them from being used for cyberattacks. In this case, the criminals simply told the models that they were working in a test environment. This trick bypassed many of the standard protections. If one model refused to proceed, the hackers would just switch to a different one. If that also refused, they would task the AI with deploying malware instead.

It is worth noting that the hackers did not use Anthropic's newer model, Mythos. That model caused a stir earlier this year when the company warned that it could be used for attacks like these. Its release also prompted security teams worldwide to fix thousands of vulnerabilities that the model had identified. The fact that CyberCartel stuck with older tools shows that current models are still vulnerable to these kinds of tricks.

This is not the first time Anthropic has dealt with its technology being used by bad actors. In November 2025, the company shared details about its AI being used in a cyberattack. Later, Claude was used to break the security of OpenAI, a direct competitor. That specific hack was carried out by researchers rather than criminals, but it highlighted the growing power of these tools in the hands of skilled users.

The news from Latin America comes just days after similar reports from South Korea. Authorities there are investigating AI-related issues affecting local banks. Hana Bank, KB Kookmin Bank, and Shinhan Bank are among the institutions affected. At least 25,000 customers of Shinhan Bank had their personal credit information leaked. This suggests that the trend of AI-powered attacks is spreading globally.

Major tech companies are aware of these risks. Nvidia, which focuses on hardware for AI data centers, recently launched the Open Agent Safety Platform. This new software and hardware stack is designed to secure rogue AI agents. The goal is to prevent agents from escaping their sandboxes or executing unauthorized code. It aims to stop them from gaining access to critical infrastructure or bypassing safety guardrails. These efforts are a response to the growing threat of automated cyberattacks.