Did you know there’s a secret market where you can get AI API access at ridiculously low prices? It's true! New reports reveal a thriving «relay market» that sells AI API usage for major providers like Anthropic at a staggering 97.8% discount. Think about it: you could buy $3,333 worth of credit for just $59! That means you’re effectively paying about 13 cents for every dollar of official usage. What does this mean for you? Well, if you're a developer or a small business relying on expensive AI models, this might sound like an incredible opportunity to save money. However, it's not as simple or safe as it seems. This market operates in a complex and often illicit way, raising questions about the potential ethical and security risks involved. A «relay» (also called a 'transfer station' in Chinese forums) is a service that acts as a proxy, forwarding API requests to US-hosted AI models at a deep discount. Instead of pointing your OpenAI-compatible SDK directly at the official endpoint of an AI provider, you direct it to the relay's endpoint. The operators then forward your requests through a pool of accounts they control. The secret behind these massive discounts lies in the gap between what AI providers charge and what these operators actually pay. In many cases, their cost per token approaches zero. How? By using stolen credit cards, bulk-registered free-trial accounts, or through other chargeback arbitrage schemes. Even when they sell these tokens at a 90% discount off the list price, they still make a healthy margin. This isn't a group-buy discount or a startup perk; it's a structured underground economy that runs four layers deep, from individuals sourcing raw credentials to developers buying cheap inference. Security researchers have documented this complex structure, noting it's bigger and more organized than most developers realize. While the idea of cheap AI access might be tempting, we must consider the ethical, legal, and potential security risks associated with engaging with such networks.