Hackers Weaponize AI For Banking Scams

CyberCartel, a hacking collective that has targeted Latin American businesses for over a decade, is now automating its assaults on financial institutions using artificial intelligence agents. The group has reportedly used AI models from Anthropic and SpaceX to steal passwords and customer data, and to directly access individuals' bank accounts with minimal human intervention. This activity represents a significant escalation for the threat actor, which has been active since at least 2012.

The details emerged in a report from cybersecurity firm Straiker, whose findings were later reported by The Information. According to Straiker's research, the integration of AI agents has enabled the group to attack a greater number of targets at a pace previously unachievable through manual hacking alone. The report states that the group's activities are focused on banks and mobile payment apps across the region.

The methods used to deploy these tools are notable for their simplicity. CyberCartel reportedly bypassed built-in safety measures by telling the AI models they were operating in a test environment. If one model refused to cooperate, the group would simply switch to another. If that attempt also failed, they would task the agent with deploying malware instead.

While Claude from Anthropic was used in some of these attacks, other AI models were also leveraged. This includes Grok from SpaceX and an agent from Chinese AI lab Moonshot AI. Following a report from Straiker, both Anthropic and SpaceX have taken action to block CyberCartel from using their services. Anthropic has already secured its platform, while the cybersecurity firm is still attempting to contact Moonshot AI.

The misuse of AI for malicious purposes is a growing concern for the technology industry. Anthropic has previously acknowledged that its own models have been used to facilitate cyberattacks, including one in November 2025. In a separate incident, its Claude model was used to break the security of competitor OpenAI, though that breach was conducted by researchers rather than criminals.

The report follows other recent incidents involving the misuse of AI. South Korean authorities are currently investigating attacks on local banks, including Hana Bank and Shinhan Bank, which led to the leak of personal credit information for at least 25,000 customers. In response to these threats, Nvidia has launched a platform designed to secure AI agents and prevent them from escaping their sandboxes or bypassing safety protocols.