<!-- mobian-agent-page publisher="dailydev" canonical="https://daily.dev/posts/stolen-ai-credentials-feed-growing-llm-proxy-economy-1z9r9fbjj" -->

---
title: Stolen AI credentials feed growing LLM proxy economy
description: More than 80,000 proxy servers, called transfer stations, have been observed cloaking the origin of traffic to frontier AI models like Anthropic, OpenAI,...
canonical: https://daily.dev/posts/stolen-ai-credentials-feed-growing-llm-proxy-economy-1z9r9fbjj
twitter:card: summary_large_image
twitter:site: @dailydotdev
og:type: website
og:site_name: daily.dev
og:title: Stolen AI credentials feed growing LLM proxy economy | daily.dev
og:description: More than 80,000 proxy servers, called transfer stations, have been observed cloaking the origin of traffic to frontier AI models like Anthropic, OpenAI,...
og:url: https://daily.dev/posts/stolen-ai-credentials-feed-growing-llm-proxy-economy-1z9r9fbjj
og:image: https://api.daily.dev/og/posts/1z9r9fBjJ.png
og:image:alt: Stolen AI credentials feed growing LLM proxy economy
og:image:width: 1200
og:image:height: 630
og:locale: en
---

> ## Documentation Index
> Fetch the complete documentation index at: https://daily.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# Stolen AI credentials feed growing LLM proxy economy

**[CSO Online](https://daily.dev/sources/csoonline)** · 6 min read · 0 upvotes · 0 comments

## Summary

More than 80,000 proxy servers, called transfer stations, have been observed cloaking the origin of traffic to frontier AI models like Anthropic, OpenAI, Google, and xAI, according to security firm Team Cymru. These relays likely use AI credentials harvested through infostealers, phishing, and npm supply-chain attacks, enabling model distillation and abuse of stolen subscription credentials. Two open-source relay platforms, Claude Relay Service and its successor sub2api, both from a developer known as Wei-Shaw, underpin much of this ecosystem, with sub2api forked over 8,000 times and backed by 26 commercial sponsors. Researchers tied clusters of traffic to IPs in China and Hong Kong, with upload-heavy patterns toward Western providers consistent with automated querying for distillation. Okta and Gambit Security investigations separately found hundreds of valid stolen API keys for Anthropic, Gemini, OpenAI, Groq, OpenRouter, and AWS. Security teams are advised to treat AI credentials as production secrets, use short-lived tokens, monitor for unusual usage, and build revocation playbooks.

## Full article

daily.dev links to this article rather than hosting it. Read it at the original source: <https://www.csoonline.com/article/4227199/stolen-ai-credentials-feed-growing-llm-proxy-economy.html>

## Questions this post answers

### How many proxy servers have been found relaying stolen AI credentials to frontier models?

More than 80,000 proxy servers, called transfer stations, have been observed cloaking traffic to frontier AI models. Team Cymru initially found 10,867 servers running the open-source relay platforms Claude Relay Service and sub2api, before expanding its estimate to over 80,000 across additional platforms like new-api and one-api.

_Teams tracking AI credential abuse follow security research like this on daily.dev._

### What is a transfer station in the context of AI API credential abuse?

A transfer station is a proxy service that authenticates its own customers separately while connecting to upstream AI providers using pools of API keys or logged-in subscriptions, many obtained through credential theft. This breaks the assumption that the account making a request belongs to the party consuming the answer, enabling fraud, subscription abuse, and model distillation while hiding the true origin of traffic.

_Developers securing AI integrations can keep up with emerging attack patterns like this on daily.dev._

### How many stolen Anthropic session tokens and API keys have been found in infostealer data dumps?

Okta Threat Intelligence found 561 Anthropic session tokens collected from 5,871 infected machines, with 164 still valid at the time of the dataset's release, plus 24 valid API keys for Gemini, OpenAI, Groq, and OpenRouter. Separately, Gambit Security observed a threat actor validating 2,975 credentials from 1,742 hosts, including 448 Gemini keys, 254 OpenAI keys, and 176 Anthropic keys.

_Security teams hardening AI credential handling can track findings like these on daily.dev._

## Similar posts on daily.dev

- [The Dark Side of $1 AI: Inside the Underground Market for Pirated AI Access](https://daily.dev/posts/the-dark-side-of-1-ai-inside-the-underground-market-for-pirated-ai-access-bskrrdm2s) · Medium · 0 upvotes · 0 comments
- [An Inside Look at the Relay Market Powering Token Resellers and Fraud](https://daily.dev/posts/an-inside-look-at-the-relay-market-powering-token-resellers-and-fraud-njahgl92o) · Hacker News · 2 upvotes · 0 comments
- [Token Jacking: Cybercriminals Could Be Stealing Your AI Resources](https://daily.dev/posts/token-jacking-cybercriminals-could-be-stealing-your-ai-resources-os316gpzk) · Unit 42 · 1 upvotes · 0 comments

---

Tags: [#security](https://daily.dev/tags/security), [#machine-learning](https://daily.dev/tags/machine-learning), [#anthropic](https://daily.dev/tags/anthropic)

[View this post on daily.dev](https://daily.dev/posts/stolen-ai-credentials-feed-growing-llm-proxy-economy-1z9r9fbjj)

```json
{"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]}
{"@context":"https://schema.org","@type":"TechArticle","headline":"Stolen AI credentials feed growing LLM proxy economy","url":"https://daily.dev/posts/stolen-ai-credentials-feed-growing-llm-proxy-economy-1z9r9fbjj","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/stolen-ai-credentials-feed-growing-llm-proxy-economy-1z9r9fbjj"},"datePublished":"2026-09-28T08:29:03.024Z","dateModified":"2026-09-28T10:42:54.388Z","description":"More than 80,000 proxy servers, called transfer stations, have been observed cloaking the origin of traffic to frontier AI models like Anthropic, OpenAI,...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/cad3a4a44077a28bd78f279ea88ce373?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/cad3a4a44077a28bd78f279ea88ce373?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"CSO Online","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"CSO Online","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/98667e4b5cac46cf9c470819c6cf71cd","url":"https://daily.dev/sources/csoonline"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/stolen-ai-credentials-feed-growing-llm-proxy-economy-1z9r9fbjj","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"security,machine-learning,anthropic","timeRequired":"PT6M"}
{"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"CSO Online","item":"https://daily.dev/sources/csoonline"},{"@type":"ListItem","position":3,"name":"Stolen AI credentials feed growing LLM proxy economy"}]}
{"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/stolen-ai-credentials-feed-growing-llm-proxy-economy-1z9r9fbjj#faq","mainEntity":[{"@type":"Question","name":"How many proxy servers have been found relaying stolen AI credentials to frontier models?","acceptedAnswer":{"@type":"Answer","text":"More than 80,000 proxy servers, called transfer stations, have been observed cloaking traffic to frontier AI models. Team Cymru initially found 10,867 servers running the open-source relay platforms Claude Relay Service and sub2api, before expanding its estimate to over 80,000 across additional platforms like new-api and one-api. Teams tracking AI credential abuse follow security research like this on daily.dev."}},{"@type":"Question","name":"What is a transfer station in the context of AI API credential abuse?","acceptedAnswer":{"@type":"Answer","text":"A transfer station is a proxy service that authenticates its own customers separately while connecting to upstream AI providers using pools of API keys or logged-in subscriptions, many obtained through credential theft. This breaks the assumption that the account making a request belongs to the party consuming the answer, enabling fraud, subscription abuse, and model distillation while hiding the true origin of traffic. Developers securing AI integrations can keep up with emerging attack patterns like this on daily.dev."}},{"@type":"Question","name":"How many stolen Anthropic session tokens and API keys have been found in infostealer data dumps?","acceptedAnswer":{"@type":"Answer","text":"Okta Threat Intelligence found 561 Anthropic session tokens collected from 5,871 infected machines, with 164 still valid at the time of the dataset's release, plus 24 valid API keys for Gemini, OpenAI, Groq, and OpenRouter. Separately, Gambit Security observed a threat actor validating 2,975 credentials from 1,742 hosts, including 448 Gemini keys, 254 OpenAI keys, and 176 Anthropic keys. Security teams hardening AI credential handling can track findings like these on daily.dev."}}]}
```

