THE CYBER BRIEF - ISSUE 52

Editor's Notes:

Economics are driving "innovation" in emerging AI gray markets.

T H I S I S S U E - S P O T L I G H T

Piggybacking to the Frontier: The Firms Selling RL Environments to Chinese Labs

• American data vendors (Mercor, Surge AI, AfterQuery, Turing) reportedly sell about $500 million in data to Chinese AI labs.
  

• RL piggybacking allows Chinese labs to improve models at the final stage of post-training, making it arguably more important than distillation. Supervised fine-tuning (SFT) is the process before RL where distillation occurs.
  

• There are dozens of data vendors, each with their own expert networks, domain specialties and data pools. A new data company sprouts up daily, trying to take advantage of the gold rush. With so much competition, many may find revenue sources in Chinese labs and even come to depend on them; one American vendor, AfterQuery, is reported to source half of its $100 million ARR from Chinese business.
  

https://www.chinatalk.media/p/the-us-is-selling-rl-environments

AI Token Abuse: What It Is and How to Prevent It

• AI tokens and credits have real resale value, and attackers know it. They create fake accounts to farm free trials, take over legitimate accounts to burn through someone else’s credits, fund usage with stolen credit cards, and sell access to anyone willing to pay.
  

• The same accessible automation that helps legitimate developers also lets fraudsters craft their attacks: They can script account creation, work around signup friction, and distribute activity across many identities at low cost. Mass account creation and credit farming that used to require real effort now run as commodity operations.
  

• Recommended defenses include scoped and rotated API keys, per-account spending limits, checks at high-value moments (signup, trial claims, purchases, key creation), binding identity to device, and monitoring velocity patterns.
  

https://fingerprint.com/blog/ai-token-abuse/

The Rise of LLMJacking: How Threat Actors Monetize Compromised Cloud and Model Resources

• The LLM Mixer Jet Night (jet-night.com) made more than 500,000 LLM API calls in under eight hours, passing nearly $1 million USD in token costs to a victim, and it advertised an unlimited plan for about $60/month, versus a typical $100–$250/month, suggesting reliance on stolen keys.
  

• Unlike the gradual resource consumption associated with cryptojacking, LLMJacking can generate large spikes in fraudulent cloud charges in a matter of hours. Across individual incidents, we observed fraudulent charges associated with resource consumption exceeding $100,000 USD in a single day, with cumulative charges at some organizations reaching $1 million USD before access was revoked.
  

• While low-level resource theft has involved the manipulation of customer support chatbots to conduct unauthorized tasks like code generation, threat actors conducting LLMJacking bypass these front-end interfaces to target backend model resources directly.
  

https://cloud.google.com/blog/topics/threat-intelligence/llmjacking-monetizing-compromised-cloud-models

Inside the underground market for AI access

• Gray-market sellers offer AI access at 70% to 90% below standard subscription prices. One case Okta cites, from Palo Alto Networks Unit 42 research, involves a stolen-access bill that reached nearly $1 million before it was contained.
  

• Okta analyzed a 7GB infostealer dump posted to Telegram on August 2, 2026, with logs from 5,871 infected machines in 162 countries. It held thousands of valid session tokens for major tech and AI providers, plus dozens of plaintext API keys.
  

• A service called “Poison Claude” sells cheap access to Anthropic models using $100 AWS bonus credits, and an exposed API route suggested nearly 900 active users. Separately, one AI video company was hit by more than 105,000 brute-force fake sign-up attempts in a single month.

https://www.okta.com/newsroom/articles/inside-the-underground-market-for-ai-access/

This newsletter is intended for informational purposes only. The content reflects publicly available information and general industry trends as of the date of publication and may not reflect the most current regulatory developments. Newsletter may contain links to vendor supported content which does not represent endorsement or promotion of products or services. Nothing herein constitutes legal, compliance, or professional advice, and should not be relied upon as such.