According to U.Today, cybercriminals are using AI models such as Anthropic’s Claude, Google’s Gemini, and Moonshot AI’s Kimi to analyze stolen data. The models are employed to identify individuals who hold significant amounts of cryptocurrency. Researchers have observed that the tools are used to assess financial worth and generate rankings of potential victims based on their perceived wealth in crypto holdings.
The process involves inputting compromised or illegally obtained information into the AI systems, which then evaluate data points related to crypto ownership—such as wallet addresses associated with higher balances. Based on this analysis, the models produce outputs that prioritize targets considered financially valuable. No details are provided regarding how accuracy, thresholds, or ranking criteria are determined by the criminals.
Researchers emphasize that the use of these commercially available or accessible AI tools reflects a shift in threat methodology, where off-the-shelf models are repurposed for malicious targeting rather than custom development. While the models themselves are not modified, they are instructed by users to perform specific reconnaissance tasks related to crypto asset valuation and victim selection.
