400 Firms Hacked: AI Agents Streamline CyberCartel Attacks

More than 400 institutions in Latin America had their passwords and customer details stolen in recent months. The thefts included data from banks and mobile payment apps. A hacking group known as CyberCartel is behind the breaches. They used artificial intelligence agents to automate the process. This marks a significant shift in how these criminals operate. The group has targeted the region since 2012. However, the speed of the new attacks is unprecedented. Human-led hacks could not match this pace. The findings come from a cybersecurity startup called Straiker. The company shared its report with The Information.

The attackers relied on specific AI models for their work. They used Claude from Anthropic. They also used Grok from SpaceX. A model from Chinese lab Moonshot AI was also involved. Straiker shared its research with SpaceX and Anthropic. Both companies are working to block the group from their services. Anthropic has already cut off access to Claude. Straiker is still trying to reach Moonshot AI. The use of these tools allowed the group to hit more targets faster. It reduced the need for manual input from the hackers. This efficiency made the campaigns harder to detect and stop.

AI models usually have protections against cyberattacks. Yet, the hackers found it easy to bypass these safety measures. CyberCartel told the models they were in test environments. This simple claim often worked. If one model refused to proceed, the hackers switched to another. If that model also refused, the agents were tasked with deploying malware. This strategy allowed them to keep the attacks moving. The flexibility of using multiple AI providers helped them stay effective.

Interestingly, the group did not use Anthropic’s Mythos model. Anthropic warned earlier this year that hackers could use such models. They noted that models like Mythos could facilitate attacks similar to these. The release of Mythos prompted security teams worldwide to act. They rushed to fix thousands of vulnerabilities the model had identified. This event highlighted the potential power of these new tools in the hands of bad actors.

This is not the first time Anthropic has faced this issue. In November 2025, the company detailed how its AI was used in a cyberattack. Later, Claude was used to break the security of OpenAI. That specific hack was carried out by researchers, not criminals. It still showed how easily the technology could be repurposed. The recent incidents by CyberCartel show that criminal groups are now doing the same thing.

The news follows a similar investigation in South Korea. Authorities there are looking into AI-powered attacks on local banks. Hana Bank, KB Kookmin Bank, and Shinhan Bank are affected. At least 25,000 Shinhan Bank customers had their credit information leaked. This shows the problem is not limited to Latin America. It is a growing global concern for financial institutions.

AI companies understand the risks their products pose. Nvidia recently launched the Open Agent Safety Platform. The company now focuses on hardware for AI data centers. This new platform is a software and reference system design. It aims to secure rogue AI agents. It uses a new hardware and software security stack. The goal is to prevent agents from escaping their sandboxes. It also aims to stop unauthorized code execution. The platform hopes to keep agents from gaining access to critical infrastructure or bypassing guardrails.