DeepSeek and Moonshot AI Used GPT-6 Astra to Train AI Models

Chinese AI developers DeepSeek and Moonshot AI actively used the results of OpenAI's GPT-6 Astra model to train their newest artificial intelligence systems, according to a report by The Information. The practice in question is distillation, which allows the capabilities of more powerful foundational models to be transferred into alternative systems. This situation highlights growing concerns regarding data acquisition and intellectual property in the tech sector.

• Utilization of GPT-6 Astra results to train Chinese AI models
• Application of distillation technique to transfer powerful AI capabilities
• Impact on recent releases including Moonshot AI's Kimi K3.1 Pro and DeepSeek V4.1-Flash 1017
• Ongoing disputes between OpenAI, Anthropic, Google, and Asian developers over data scraping

The Role of Distillation in AI Development According to Reuters, the distillation results from GPT-6 Astra were leveraged in training Moonshot AI's Kimi K3.1 Pro model released last week, as well as the updated DeepSeek V4.1-Flash 1017 which appeared on October 17. Distillation involves using the outputs of a more capable model to train another system, enabling developers to replicate advanced features faster and without comparable research costs. Major AI firms argue that this practice enables competitors to copy sophisticated capabilities without making equivalent financial or computational investments. OpenAI previously reported discovering data collection campaigns linked to Moonshot AI starting back in July.

Industry Response and Security Measures The controversy underscores tightening restrictions and security measures across the artificial intelligence industry. Earlier in October, Anthropic announced that new defensive protections implemented for its Claude models successfully reduced the volume of data gathered via distillation attacks by 90 to 99 percent. Representatives for OpenAI, DeepSeek, and Moonshot AI have not yet responded to requests for comment from major news outlets. As regulatory scrutiny increases, the debate over model distillation and fair competition is set to remain a critical issue for global technology developers.