The Dark Side of the Moon · Full chronicle and history

On October 28, The Information reported that the latest models from DeepSeek and Moonshot AI are suspected of using output from OpenAI's GPT-6 Astra for distillation training. In September, OpenAI traced distillation activities that began in July to individuals associated with Moonshot AI. Anthropic stated that it has reduced the data acquisition for distillation attacks against Claude by approximately 90% to 99%.

A user on X revealed that after refusing to cooperate and criticizing a certain creator for spreading propaganda content, Kimi AI again sent a private message asking them to delete their comment; the user refused and expressed anger.

On October 24, X blogger Elshayib posted what he said were promotional documents from Kimi (Moonshot AI), requiring bloggers promoting K3.1 Pro to submit drafts for review first, not to label paid partnerships, and providing talking points such as the AI gap between China and the US having narrowed to one month. He refused to participate and struck back by posting screenshots of private messages.

On October 12, Claude developer Anthropic released a follow-up article naming Zhipu and Moonshot, stating that it had banned tens of thousands of accounts related to distillation attacks and rejected trillions of tokens of paid traffic. It estimates that the new defenses have reduced the volume of extracted inference data by approximately 90% to 99%.

On September 11, 2026, the US AI giant Anthropic released a long-form investigative report detailing how Chinese AI companies such as Moonshot AI and DeepSeek, while distilling own models from US models like Claude, allegedly sent real internal prompts and sensitive data of users from the People's Liberation Army, police, and state-owned enterprises to US AI platforms, sparking strong concern over model distillation security and data leakage.

On September 10, Anthropic released a threat intelligence report stating that Chinese AI companies, including DeepSeek and Moonshot, used third-party API proxy services to bypass regional restrictions and call Claude. The report also noted that these relay services may obtain model access through methods such as fake accounts and stolen API keys.

On August 8, 2026, British media disclosed that domestic large-model company Moonshot AI is adjusting its equity structure and introducing multiple state-backed investors, and has completed its conversion to a joint-stock company, in preparation for advancing plans to list in Hong Kong.

According to a report by the Financial Times on August 8, Chinese AI company Moonshot AI is adjusting its equity structure and introducing multiple state-backed investors to prepare for a future IPO in Hong Kong.

The U.S. Department of the Treasury has warned that Moonshot AI's project, Kimi K3, may be stealing U.S. AI models through "distillation" techniques and could potentially use overseas nodes to hijack access to NVIDIA servers. The U.S. government has added Kimi K3 to the Entity List for sanctions.

On July 22, White House officials accused Chinese artificial intelligence company Moonshot AI of illegally distilling Anthropic's Fable model to develop the K3 system, and of obtaining banned Nvidia chips.

Chinese AI startup Moonshot AI officially released its new large model, KIMI K3, on July 17.

Beijing's Moonshot AI has released the Kimi K3 model, with a total scale of 2.8 trillion parameters, capable of processing 1 million tokens, and achieving leading results in tests such as the Arena.