Researchers Uncover 'LLMjacking' Scheme Targeting Cloud-Hosted AI Models

**LLMjacking:** Researchers observed attackers exploiting a Laravel vulnerability to exfiltrate cloud credentials and gain access to cloud-hosted LLM services (e.g., Anthropic Claude), using key-checkers and reverse-proxy tooling to validate and monetize access while attempting to evade logging; the abuse can generate substantial ongoing costs for victims and disrupt legitimate use.