According to CoinMeta, OpenAI claimed to have shut down an activity that attempted to replicate the reasoning mechanism of their AI model. The core group involved a Chinese startup company, Moonshot AI, which developed the Kimi chatbot. The activity began on July 1st, and OpenAI recorded 16,000 extraction requests on July 24th and 25th alone, involving over 4,000 users. OpenAI stated that the activity was completely halted on July 28th, with the goal of replicating the reasoning process rather than the answers. Modern AI models engage in gradual reasoning before responding, and OpenAI ensures that this process remains encrypted, as extracting information could reveal details of the final answers that were intended to be hidden. OpenAI pointed out that the operators did not break the encryption or directly access user conversations; instead, they manipulated the model interactions to reappear the protected reasoning in a visible form, which violated the service terms. OpenAI has closed a specific pathway to prevent the use of others' encrypted reasoning replays for restoring content.