The article discusses the rise of LLMjacking, a type of cybercrime where hackers hijack login credentials to access large language models (LLMs), exploiting them for malicious purposes. Researchers at Sysdig highlight how, within hours, accounts such as those from DeepSeek can be commandeered, leading to unauthorized generation of content and circumvention of regional restrictions. Techniques like OAI reverse proxies (ORPs) provide hackers with anonymity, making it harder for providers to detect the abuse. The trend is alarmingly visible in regions with strict LLM access regulations, with implications for all AI provider account holders.
"LLMjacking, the act of hijacking accounts connected to large language models, is increasingly becoming a form of cybercrime that bypasses traditional usage restrictions."
"Research shows that hackers target login credentials for LLMs, exploit stolen data through a reverse proxy, and carry out various malicious activities at the original user's expense."
Collection
[
|
...
]