LLM Hijackers Quickly Incorporate DeepSeek API Keys

A new cybercrime called "LLMjacking" involves stealing credentials to access expensive AI language models like GPT-4. Attackers use reverse proxies to hide their activities, generating content or creating malicious scripts without paying. Some spread usage across multiple stolen accounts, while others cause individual users to incur large unexpected charges. This trend emphasizes the need for improved security and cost monitoring for AI services.