AI's Dark Side: Hackers Exploit Models for Crime

The cyber criminal underworld is increasingly prioritizing illicit access to AI models and powerful computing resources. Hackers are actively engaged in 'LLM-jacking,' which involves stealing credentials for public AI tools and hijacking computing power to run their own models for free. This burgeoning underground economy offers AI access at heavily discounted rates, with some marketplaces on the dark web providing up to 97% off subscriptions for advanced AI models.

This cheaper access provides cyber attackers with a significant financial and efficiency advantage. They are leveraging these stolen AI capabilities for various malicious activities, including extortion, warfare, and espionage. State-backed groups, such as a Chinese cyber espionage unit, have been observed exploiting cloud-hosted servers to run their own AI models, mirroring tactics used for cryptocurrency mining.

Experts warn that any organization viewing AI as a temporary trend risks facing an overwhelming surge in cyber incidents. As more corporations host customized AI models on-premises, these systems will also become prime targets. Attackers can exploit the increased compute usage associated with new AI infrastructure to hide their malicious activities, making current company systems vulnerable.