Technology & AI

DeepSeek, Moonshot AI reportedly trained models on OpenAI's Astra

Chinese AI labs turned to GPT-6 Astra, OpenAI's most powerful model, after Anthropic tightened its defenses.

By Standard Topic. AI-generated from the reporting listed under Coverage. Published . Updated .

This story is no longer being updated. It reflects reporting up to .

What happened

DeepSeek and Moonshot AI reportedly used outputs from OpenAI's GPT-6 Astra to train their latest models, according to The Information. Chinese AI labs have increasingly shifted to Astra as a distillation target after Anthropic tightened its defenses.

Why it matters

The reports have renewed concerns about distillation, a technique that can help competitors reproduce advanced AI capabilities.

Timeline

  1. Oct 29, 9:20 AM ET: Chinese labs were mainly tapping OpenAI's GPT-6 Astra instead of Anthropic's Claude after Anthropic tightened its defenses.
  2. Oct 29, 5:05 PM ET: DeepSeek and Moonshot AI were reportedly trained on Astra outputs for their latest releases.
  3. Oct 30, 9:13 AM ET: The Information's account renewed concerns over AI distillation, a technique that can help competitors reproduce advanced AI capabilities.

Coverage

More from Technology & AI

All Technology & AI news · Technology & AI story archive