Matt Lenhard wrote that scammers were selling access to advanced artificial intelligence models for one-tenth of the market price. The method relies on obtaining tokens for free or below cost and marketing them through a relay API.
Scammers are exhausting the free credits of new accounts, using invalid payment methods, or reversing payments after high usage with temporary cards. They are also reverse-engineering AI-enabled consumer software and exploiting its back ends, while targeting poorly coded support chatbots.
According to Lenhard, the way to reduce this abuse is to add friction to transactions and increase the scammers’ costs. As this model, which is built on obtaining tokens for free, becomes more difficult, scammers are expected to turn to easier targets.
Why it matters
This method shows that the gap between the apparent price of access to AI models and the actual cost of providing that access can be exploited. Because free credits, payment systems, and the back ends of consumer software can all be used in the same chain, the issue is not limited to the security of a single account type; it requires model providers to monitor different access channels together. Additional verification and transaction steps may increase the cost of fraud, but the extent to which these steps would make access more difficult for legitimate users is left open in the text. For this reason, the focus of the debate is not only the sale of cheap tokens, but also the balance between ease of access and preventing abuse.