Hackers Steal AI Platform API Keys to Resell Model Access
Attacks on roughly 10 AI development platforms including Railway led to stolen keys and up to $1 million in charges per victim, according to the WSJ.
Over the past month, about 10 AI development platforms, including Railway, were hit by cyberattacks in which hackers targeted API credentials, according to reporting by the WSJ. In the Railway incident, attackers stole $371,000 worth of cryptocurrency along with Railway customers' API keys.
The stolen keys generated at least $750,000 in usage charges across AI services including Amazon Bedrock, OpenAI, Anthropic and Google Gemini, the WSJ reported. AI infrastructure company Zeabur also suffered an API key theft that resulted in $100,000 in charges.
Google Threat Intelligence Group (GTIG) and Mandiant describe the theft of corporate AI model access for use or resale as "LLMJacking." Recent attacks targeted financial services, retail, technology and transportation companies in North America, South America and the Asia-Pacific region.
Resale market for stolen model access
One attacker made more than 500,000 LLM API calls in under 8 hours, creating and deleting short-lived service accounts to evade rate limits. A single illegal token routing service offered unlimited frontier model access for about $60 a month while the victim company was billed $1 million in model usage.
Chainalysis found more than 700 such brokerage services online, with the top five taking in a combined $6 million in cryptocurrency. Attacks on cloud-based AI development platforms targeting AI model access began in early October 2026 and show no sign of letting up.
Quick answers
What is LLMJacking?
It is the term used by Google Threat Intelligence Group and Mandiant for attackers stealing corporate AI model access and using or reselling it.
How much have victims been charged?
Stolen Railway keys led to at least $750,000 in usage charges across AI services, and a single illegal token routing service was linked to $1 million in model usage billed to a victim company.
Which companies were affected?
About 10 AI development platforms were hit, including Railway, whose customers' API keys were stolen, and AI infrastructure company Zeabur, which saw $100,000 in charges.