中国AI实验室转向OpenAI的GPT-6 Astra，因Anthropic加强防御
中国AI实验室现在以OpenAI的GPT-6 Astra模型训练自己的模型，如Kimi K3.1 Pro和V4.1-Flash 1017，因为Anthropic加强了防御。这影响了AI模型开发。
chinese ai labs, including deepseek and moonshot ai, have shifted their focus to openai's gpt-6 astra model to train their own models, such as kimi k3.1 pro and v4.1-flash 1017. this change occurred after anthropic implemented new methods that reduced data extraction by 90 to 99 percent, making it harder for labs to access its claude model's outputs. openai reported a coordinated distillation campaign in late september, involving individuals linked to moonshot ai. openai takes the problem seriously and continues to roll out new defenses.