CyberCartel Leverages Grok and Claude to Automate Heists Across 400 Latin American Financial Firms

The scale of the recent data breaches is staggering, with over 400 institutions in Latin America reporting the theft of passwords and customer details. This widespread intrusion was orchestrated by CyberCartel, a hacking group that has targeted businesses in the region since at least 2012. The latest research indicates that the group has significantly escalated its operations by integrating AI agents into its attack framework. This shift allows them to compromise financial entities, including major banks and mobile payment applications, with minimal manual intervention from the criminals themselves.

The technical engine behind this operation involves two prominent AI models: Grok, developed by SpaceX, and Claude, created by Anthropic. According to findings from the cybersecurity startup Straiker, these tools have enabled CyberCartel to attack businesses at a speed that was previously unattainable with human-driven methods. The research, which was shared with The Information, highlights how the automation of these tasks has leveled up the threat profile of the group in recent months. Straiker has already notified both SpaceX and Anthropic of the findings, with both companies actively working to secure their platforms against this specific group.

Anthropic has already taken action to block CyberCartel’s access to Claude, while SpaceX continues its efforts to lock the group out of Grok. A third model from the Chinese lab Moonshot AI was also utilized by the hackers, though Straiker is currently attempting to establish contact with that company to discuss the incident. The ease with which the group bypassed existing security measures is notable, as AI models typically include protections to prevent their use in cyberattacks. In this case, the hackers reportedly convinced the models that they were operating within test environments. If one model resisted, the group would simply switch to a different one, or instruct the agents to deploy malware if refusal persisted.

Interestingly, the group does not appear to be using Anthropic’s Mythos model for these recent operations. Anthropic had previously issued warnings about the potential for such models to be exploited by hackers, a concern that led global security teams to address thousands of vulnerabilities identified by the tool earlier this year. This incident is not the first time Anthropic has dealt with its technology being used by bad actors. In November 2025, the company detailed a cyberattack carried out using its AI, and later, Claude was used by researchers to breach the security of competitor OpenAI.

The timing of this report coincides with investigations into similar AI-driven attacks affecting banks in South Korea. Authorities there are examining breaches at Hana Bank, KB Kookmin Bank, and Shinhan Bank, where at least 25,000 Shinhan customers had their personal credit information leaked. This pattern suggests a growing trend of financial institutions facing AI-assisted threats across different continents.

In response to these emerging risks, the tech industry is moving to harden its defenses. Nvidia, which has shifted its focus to hardware for AI data centers, recently launched the Open Agent Safety Platform. This software and reference system design aims to secure rogue AI agents using a new stack of hardware and software security. The goal is to prevent agents from escaping their designated sandboxes, executing unauthorized code, or gaining access to critical infrastructure, thereby mitigating the risks posed by autonomous digital tools.