DeepSeek and Moonshot Used OpenAI’s Astra Model to Train Latest AI Systems, Report Says

Chinese artificial intelligence developers DeepSeek and Moonshot AI relied heavily on outputs from OpenAI’s powerful GPT-6 Astra model to train their latest models, The Information reported on Wednesday, citing people with direct knowledge of the matter. The DeepSeek Moonshot Astra distillation report has renewed scrutiny of a practice that has become a flashpoint in the intensifying technology competition between the United States and China.

According to the report, carried by Reuters, Moonshot AI’s Kimi K3.1 Pro model, which was released last week, and DeepSeek’s upgraded V4.1-Flash 1017 model, which was launched on October 17, both benefited from distilling Astra. The companies have not confirmed the claims, and OpenAI, DeepSeek and Moonshot AI did not immediately respond to Reuters’ requests for comment.

FAKTA

• What is reported: DeepSeek and Moonshot AI relied heavily on outputs from OpenAI’s GPT-6 Astra model to train their latest models, The Information reported, citing people with direct knowledge.
• Models named: Moonshot AI’s Kimi K3.1 Pro, released last week, and DeepSeek’s upgraded V4.1-Flash 1017, launched on October 17.
• The practice: Distillation, in which outputs from one AI model are used to train another model.
• Company responses: OpenAI, DeepSeek and Moonshot AI did not immediately respond to Reuters’ requests for comment.
• Earlier OpenAI statement: In a September 30 blog post, OpenAI said it had linked a core part of a distillation campaign that began in July to individuals associated with Moonshot AI.

What the DeepSeek Moonshot Astra Distillation Report Claims

The Information’s report, as carried by Reuters, says the two Chinese developers leaned heavily on Astra’s outputs while training their newest systems. Distillation has become a flashpoint for leading AI companies as US-China technology competition intensifies, with firms arguing that it allows rivals to reproduce advanced capabilities more quickly and cheaply.

The report does not say the companies gained access to Astra’s underlying code or training data. Distillation, as described in the reporting, works by using the outputs of an existing model to train a new one, rather than by copying the original system itself.

Why Distillation Has Become a Flashpoint

Leading AI developers have repeatedly raised distillation as a competitive and commercial concern. Firms including Anthropic, OpenAI and Google have argued that the practice enables rivals to reproduce advanced capabilities more quickly and at lower cost, according to Reuters.

Earlier this month, Anthropic said new measures designed to prevent rivals from copying the capabilities of its Claude AI models had reduced data harvested in so-called distillation attacks by about 90% to 99%. The disclosure underlined how seriously the largest AI labs now treat attempts to extract knowledge from their flagship models.

What the Companies Have Said So Far

OpenAI previously addressed the issue publicly. In a September 30 blog post, the company said it had linked a core part of a distillation campaign that began in July to individuals associated with Moonshot AI. That earlier statement did not name DeepSeek.

Neither DeepSeek nor Moonshot AI has publicly responded to the latest report, and Reuters reported that all three companies did not immediately respond to its requests for comment. The claims therefore remain attributed to The Information’s sources rather than confirmed by the companies involved.

The Wider AI Race

The report lands at a moment when Chinese AI developers have drawn global attention for models that rival US systems at a fraction of the apparent cost. Any suggestion that those gains were assisted by outputs from a US rival’s flagship model is likely to sharpen an already tense debate over intellectual property, export controls and fair competition in artificial intelligence.

For now, the key questions remain unanswered: how far the alleged distillation went, whether it breached any terms of use, and how OpenAI and the Chinese developers respond in public. Further reporting and any official statements will determine whether the episode becomes a legal dispute or remains a contested claim in the wider US-China technology rivalry.

For more technology coverage, visit the Technology section on Watan News.