LLMjacking: From Emerging Threat to Black Market Reality
Blog post from Sysdig
LLMjacking, which emerged in May 2024 as a novel security threat, has evolved into a commercialized cybercrime marketplace by early 2026, paralleling the growth of cryptomining. This type of attack involves the unauthorized use of cloud-hosted Large Language Model (LLM) resources through compromised credentials, APIs, or exposed endpoints, leading to inflated cloud bills and potential exposure of sensitive model capabilities. Initially a theoretical concern, LLMjacking has developed into an organized ecosystem dubbed "Operation Bizarre Bazaar," where attackers monetize unauthorized AI access and sell it via underground marketplaces on platforms like Telegram and Discord. The risk to AI systems extends beyond financial costs to significant operational security threats, as attacks on Model Context Protocol (MCP) servers could lead to broader infrastructure compromises. For security leaders, this underscores the importance of robust credential management, an assume-breach mindset, and vigilant monitoring of APIs and AI integrations to mitigate the expanding risk landscape.