OpenAI Blocks Reasoning Theft – Azure Still Leaks

🔥 What happened OpenAI shut down a coordinated campaign to steal model reasoning: 16,000 requests from 4,000+ users on July 24–25, tied to 15,000+ accounts. The trail leads to people linked to Moonshot AI, maker of Kimi. 💡 Why it matters The same attack still works on Microsoft Azure — against every OpenAI model tested, including GPT-6 Astra, and Anthropic up to Sonnet 5. A single request extracts reasoning verbatim. OpenAI's own API is locked down; the cloud isn't. ⚡ Our take Hosting reasoning models without equivalent protections is an open barn door for model theft. The researchers are right: no cloud should serve reasoning models without matching safeguards — otherwise export controls are just theater.