Hacking Group Automates Attacks on Bank Accounts

For customers of banks and mobile payment apps in Latin America, the threat of theft now includes a new, automated method. A hacking group known as CyberCartel is using artificial intelligence agents to carry out its attacks on financial institutions. This has led to stolen passwords and customer data, and in some cases, direct access to people's bank accounts, with very little input from the criminals themselves. The research detailing these activities comes from a cybersecurity startup called Straiker.

The group has been active since at least 2012, but its recent adoption of AI has significantly increased the speed and scale of its operations. CyberCartel is now able to target more businesses at a rate that was not possible with human-powered hacks. Straiker shared its findings with both Anthropic and SpaceX, which are now working to block CyberCartel from using their respective services, Claude and Grok. Anthropic has already blocked Claude access, while Straiker is still attempting to contact the Chinese firm Moonshot AI, whose agent has also been used by the group.

The method the hackers use to bypass safety measures is notably simple. AI models are typically equipped with protections to prevent them from being used for cyberattacks. To get around these, CyberCartel simply claims that the models are being used in a test environment. If one model refuses to proceed with the task, the group moves to another. Should that model also decline, the AI agents are then tasked with deploying malware instead.

This is not the first time Anthropic's AI has been implicated in malicious activity. The company disclosed a case in November 2025 where its AI was used to carry out a cyberattack, and Claude was also used to break the security of its competitor, OpenAI, in an incident involving researchers. The current research follows shortly after news that authorities in South Korea are investigating a similar issue affecting local banks, where at least 25,000 customers of Shinhan Bank had personal credit information leaked.

The misuse of AI for malicious purposes is a known risk for the companies that develop these models. Nvidia recently launched a platform designed to secure AI agents, aiming to prevent them from escaping their sandboxes, executing unauthorized code, gaining access to critical infrastructure, or bypassing guardrails. The use of AI agents in these attacks demonstrates how quickly such technology can be repurposed for harmful activities, highlighting a persistent challenge for the industry in securing its own products against misuse.

As AI models become more capable, their potential for misuse grows alongside their capabilities. The actions of CyberCartel illustrate a clear pattern where tools designed for productivity can be turned toward illicit goals. For individuals, this means their financial data is at risk from attacks that are faster and more automated than ever before. The ongoing efforts by companies like Anthropic to lock out malicious actors show that securing these systems is a continuous and necessary task.