OpenAI's Astra becomes the main target for Chinese AI labs after Anthropic tightens its defenses
Summary
- Chinese AI labs including DeepSeek and Moonshot AI are increasingly targeting OpenAI's GPT-6 Astra model for model distillation.
- The labs reportedly relied on Astra's outputs to train new models such as Kimi K3.1 Pro and V4.1-Flash 1017.
- Anthropic previously served as the primary target for distillation, but tightened defenses that cut data extraction by 90 to 99 percent.
- OpenAI reported a coordinated distillation campaign in late September linked to Moonshot AI actors.
Why it matters
Model distillation from frontier models like GPT-6 Astra highlights ongoing security challenges in preventing unauthorized training data extraction by competing AI labs.