I'd argue that, given a lot of this is reselling unused capacity from paid Claude Max plans, this isn't fraud.
It's actually extremely common in China: there is a very vibrant ecosystem to resell the % of your subscriptions that you haven't used. Something we should absolutely do more of in the West.
I actually pay for the Claude Max plan myself (alongside way too many AI services but that's another issue😅) and, I just checked, I virtually never use more than 10% on a monthly basis. The reason I pay for it is that, from time to time, I need a big surge of it on a given day and given daily session limits, only the Max plan allows for it.
If I could resell the 90% of my subscription that sit unused on any given month, why the hell not? That's not defrauding Anthropic: in fact in a very real way it's Anthropic defrauding **ME** by having this setup where I'm effectively forced to pay for something I only use 10% of!
When I'm in China, to buy almost anything, I will always check marketplaces like Xianyu (闲鱼) where you can buy these unused subscriptions.
For example, every year I go do a comprehensive health check with a company called iKang (爱康) which is owned by Alibaba. If I were to buy their health check package at the rack rate, I'd pay around RMB8,000 But on these platforms, *for the exact same package*, I can pay only RMB1,000 -> 1/8th of the price!
How is this possible? Well because you have plenty of large companies out there that buy iKang subscriptions for their employees as part of their remuneration packages. A big proportion of these employees - often young and healthy - don't want to spend a full day doing dozens of health checks so they resell their packages at a heavy discount. iKang doesn't care: the packages are paid for anyhow... They're not getting "defrauded".
Same thing for hotels: many credit card companies or banks offer people free hotel stays as incentives. Many people resell those at unbelievable discounts.
As for the part on subsidizing access in exchange for user data, I'd argue that - as a user - it's also a pretty sweet deal: if I could pay $20 for Claude Max instead of $200 in exchange for agreeing that my data - on top of training Claude itself - is also used to train another open source AI, why the hell not?
That's literally the deal Google has offered you for the last 20 years. Nobody is calling Gmail fraud 🤷
Does all of this violate Anthropic's terms in some way? Probably. But Anthropic's terms also say absolutely insane stuff like the fact that if you use Claude to help you criticize Anthropic publicly, you've pre-agreed to pay their lawyers to go after you and you've pre-agreed you've lost the case (https://x.com/RnaudBertrand/status/2064892380701237647?s=20). So "it violates the ToS" isn't exactly such a moral high ground.
So, no, this isn't fraud. It's what an efficient market actually looks like.
TIL
