There is increasing competition in the global AI industry as top technology companies race to create increasingly powerful AI. Two notable AI companies in China, DeepSeek and Moonshot AI, have reportedly used outputs from OpenAI’s GPT-6 Astra to improve their newest AI systems.

Both companies heavily used data produced by OpenAI’s GPT-6 Astra for their models using AI model distillation, raising questions around ownership of AI technology and issues regarding data protection, IP and increased competition in AI between the United States and China.

GPT-6 Astra Technology Used By DeepSeek and Moonshot AI

• New models, Kimi K3.1 Pro by Moonshot AI and an upgraded version of DeepSeek called V4.1-Flash 1017, reportedly used data from OpenAI’s GPT-6 Astra.
• The Kimi K3.1 Pro was released by Moonshot AI recently. DeepSeek’s upgraded model was released on October 17th 2026. Both companies used outputs produced by the GPT-6 Astra while developing their models, sources close to the matter told The Information.
• This has raised concerns over how much advanced AI models are being used during competitive AI development.

What Is AI Model Distillation?

AI model distillation is a form of machine learning whereby a smaller or different model learns from the outputs of a more advanced model.

Instead of trying to create all the features in a model from scratch, developers can train a new model to replicate the same outputs using a powerful AI system. This can save time, computational power and money.

While AI model distillation is common practice in the technology industry, there can be issues when companies use outputs from another company’s AI model without permission.

OpenAI and other top AI developers have raised concerns that they can be used by competitors to copy advanced features.

OpenAI Raise Concerns Around AI Data Extraction

OpenAI have previously raised concerns around AI model scraping.

In a blog post released on September 30th 2026, OpenAI stated that they were able to trace a large portion of a wider model distillation effort, which started in July, to people linked to Moonshot AI.

They have expressed concern over attempts to extract advanced features and valuable reasoning from their models.

This has become a key concern for AI companies as models are often made up of years of research, investment into infrastructure and development.

Anthropic Add Additional Security Features

• OpenAI are not the only company taking steps to prevent copying of their models. Anthropic have added extra security to their Claude AI models.
• Anthropic state that their new security features have helped reduce the amount of data extracted through model distillation by around 90%-99%.
• These improvements help detect unusual automated activity, prevent large scale data extraction and make it harder for outside developers to replicate the features of proprietary AI models.
• This shows how more companies are working to improve security of their models.

AI Competition Rising Between US and China

The fact that Chinese AI developers have used outputs from GPT-6 Astra shows how competition between American and Chinese technology companies is increasing.

We have many top AI developers in the United States such as OpenAI, Anthropic and Google. However, we are also seeing new and established companies in China such as DeepSeek and Moonshot AI.

Chinese AI companies have been turning heads due to their powerful AI models and efficient training methods. However, the increased use of AI model distillation has caused debate around innovation, IP rights, cost of creating AI and fair competition within the global tech industry.

What Does This Mean For The Future Of AI?

This could impact how big tech companies secure their AI.

As AI develops further, companies will likely invest more into improving their security features, adding improved monitoring tools and policies to prevent others from extracting their proprietary model features.

This controversy shows how important it will be to find a balance between innovation and collaboration while also ensuring data is protected, AI is developed responsibly and IP rights are respected.

With OpenAI, DeepSeek, Moonshot AI and Anthropic all competing, the global AI race looks set to increase further in the coming years.