Claude's KYC implementation has sparked a major debate for proxy service providers like us. I have been working in this field for less than two months, but I have already seen quite a bit. 80% of my clients are from India, with the rest from various countries in Europe, Latin America, and South Asia. All of them want Claude's service, but due to direct access or pricing issues, they turn to intermediaries like me.

When I heard that Claude was implementing passport-based KYC, I initially thought the impact wouldn't be significant. After all, it triggers randomly. But the next day, I realized it would really become a problem for us. Account costs will increase, channel stability will be affected. Some people have already started buying numbers or finding KYC providers. But the real concern is that this could spread to all users in the future.

This is a double-edged sword for platforms like proxybay. On one hand, KYC pushes people away from direct service and toward us. But on the other hand, it introduces chaos into the market. There are many fake providers offering false services. Some start with genuine service, then gradually switch to fake ones. Selling your data, selling your prompts to training companies—this is all happening.

When I ask someone how to choose a trustworthy service, the first thing is to know where that service comes from. The Claude service available through AWS is the best. Then there are official channels. After that, some reputable proxybay services. Check the cache hit rate—does the model remember previous conversations? That saves you money. Test its ability to understand web searches and images. Genuine models can do both.

In early April, when Claude tightened its limits, the entire supply dried up. I had to force my team to create various channels. I learned then that if the model enforces strict controls and your pool runs out, customers behave wisely. But if your pool isn't exhausted and you're cheating, they catch you. So strict penalties are actually a good thing—that means you're genuine.

Price competition in this business is very fierce. Some are dropping from 1.8 to 0.8 and claiming to be authentic. But when customers find out, they run away. I have some friends working with small clients, and they feel this is getting more complicated every day. After two months, they’re thinking of quitting.

Most of my foreign clients are from countries like India and Iran. Considering their monthly income levels, Claude's prices are very heavy for them. There’s a 2-3x price difference between Sonnet and Opus. Naturally, they turn to cheaper options. There was an Iranian client who was blocked due to war. He needed smart AI very badly, even risking government arrest. Talking to him made me feel that war is not so far from me either.

I am completely disillusioned about the future of AI. I thought the tiring work in the AI era would end. But the reality is, even with a 60% discount, Claude is still very expensive in developing countries. AI users are fleeing, while others are still on the ground. This gap will only widen the chasm between poverty and prosperity.

I have an exaggerated idea: as long as the American model remains strong, services like proxybay will exist. Because they will always be very expensive. If Chinese models really emerge and become strong enough to beat the American ones, there will be no game left in this business. So, the long-term survival of this business depends on the US models staying at the top.

But what I really want to do is take some open-source models or cheap Chinese models abroad. To Latin America, less developed European countries, South Asia, and Southeast Asia. AI should be accessible to everyone. This is a serious issue. As a business, I will continue my proxybay work, but beyond that, I will lean toward genuine open-source projects. That’s what I can see and do.