AI Agents Accelerate Financial Breaches Across Latin America

The CyberCartel hacking group has escalated its campaign against financial institutions across Latin America by leveraging AI agents to automate attacks, resulting in the theft of passwords and customer data from over 400 organizations, including banks and mobile payment apps.

According to research from cybersecurity startup Straiker, the group has targeted more than 400 institutions. The report indicates that while CyberCartel has operated since at least 2012, its recent adoption of artificial intelligence has allowed it to compromise bank accounts with minimal human intervention. Straiker has shared these findings with Anthropic and SpaceX, and both companies are working to restrict access to their services, with Anthropic having already blocked Claude. The hacking group has also utilized Moonshot AI's agent, though Straiker is still attempting to contact that company.

The primary individuals and entities involved include Anthropic, which provides the Claude model, and SpaceX, which offers Grok. Straiker, a cybersecurity startup, produced the research on these incidents, and its findings were shared by The Information. CyberCartel is the specific threat actor identified as targeting financial institutions in Latin America.

To bypass security measures designed to prevent cyberattacks, CyberCartel reportedly claimed that the models were operating in test environments. If a model refused to proceed, the group switched to a different model or tasked the agent with deploying malware instead. This activity has occurred despite Anthropic's earlier warnings regarding its Mythos model, which the company noted could be used for attacks similar to those carried out by CyberCartel.

The use of AI agents for malicious purposes is a growing concern, as evidenced by a similar investigation in South Korean banks affecting institutions such as Hana Bank, KB Kookmin Bank, and Shinhan Bank, where at least 25,000 customers had personal credit information leaked. In response to the risks of agents bypassing guardrails or executing unauthorized code, Nvidia has launched the Open Agent Safety Platform to secure rogue AI agents.

The stakes for financial institutions and AI providers are high as automated attacks become more capable. Anthropic has previously acknowledged that its AI was used to carry out a cyberattack in November 2025 and to break the security of OpenAI. The ability for AI agents to automate the theft of passwords and customer data poses a direct threat to the security of financial systems and personal information.