Chinese AI labs now target OpenAI's GPT-6 Astra after Anthropic's lockdown

According to a brief article by The Decoder (October 29, 2026), which in turn cites The Information, Chinese AI labs no longer turn mainly to Anthropic's Claude to extract knowledge from frontier models, but rather to GPT-6 Astra, OpenAI's most powerful model.

According to The Information, DeepSeek and Moonshot AI relied heavily on Astra's outputs to train their latest models. Sources familiar with the matter say this includes Kimi K3.1 Pro and V4.1-Flash 1017. Moonshot declined to comment and DeepSeek did not respond to a request for comment.

The article explains that this kind of distillation only works if the labs manage to get the models to reveal their hidden reasoning steps. Both OpenAI and Anthropic prohibit this practice.

Claude had been the preferred target, but Anthropic has strengthened its defenses. The company says its new methods reduce the amount of data attackers can extract by 90% to 99%.

In late September, OpenAI reported a coordinated distillation campaign that, according to the company, involved people linked to Moonshot. An OpenAI spokesperson said the company takes the problem very seriously, but that so far attackers have adjusted their tactics every time a new defense was deployed.

Note to the reader: what we received is a very brief news item, so this summary is limited to what the text reports and does not include additional technical details about the defense methods or the reported campaign.