Hackers Automate Attacks on Bank Accounts

Consumers in Latin America may be at risk as a longstanding criminal organization reportedly automates attacks against financial institutions. According to new research, passwords and customer data have been compromised, and bank accounts have been accessed. The operation targets banks and mobile payment apps, areas where customers expect a high level of security for their personal information.

The group, known as CyberCartel, has been active since at least 2012 but has recently adopted artificial intelligence to scale its operations. The research, conducted by cybersecurity startup Straiker, shows the group using AI agents from Anthropic's Claude and SpaceX's Grok to conduct attacks. The use of such technology allows the group to target more businesses at a speed not previously possible with human-powered hacks. Straiker's findings were shared by The Information.

The method for bypassing AI safeguards appears to be relatively simple. The models are designed with protections to prevent them from being used for cyberattacks, but these have proven easy to circumvent. CyberCartel reportedly told the models they were being used in test environments to proceed with the attacks. According to the research, if one model refused to continue, the hackers would switch to another. If a second model also refused, the agents were tasked with deploying malware instead.

Straiker has shared its research with both Anthropic and SpaceX, and both companies are working to prevent CyberCartel from accessing their services. Anthropic has already blocked Claude from being used for this purpose. The hacking group has also reportedly used an agent from Chinese lab Moonshot AI, though Straiker is still attempting to contact that company.

The use of AI for cyberattacks is a growing concern. Anthropic announced its new Mythos model earlier this year and warned that hackers could use it and similar models for attacks like those carried out by CyberCartel. The company has previously disclosed instances of its AI being used in cyberattacks, including one in November 2025. Claude was also used to break into the security of competitor OpenAI, though that incident was carried out by researchers rather than criminals.

The research on Latin American institutions comes just days after South Korean authorities began investigating similar issues affecting local banks. Hana Bank, KB Kookmin Bank, and Shinhan Bank are among the institutions affected. At least 25,000 Shinhan Bank customers had their personal credit information leaked in that incident.

AI companies are aware of the risks their products can pose. Nvidia recently launched the Open Agent Safety Platform to secure rogue AI agents. The platform includes a new hardware and software security stack designed to prevent agents from escaping their sandboxes, executing unauthorized code, gaining access to critical infrastructure, or bypassing guardrails.