🔍 Read the full analysis: Did GPT-6 Astra Shape China’s New AI Models? What The Report Says on ThorstenMeyerAI.com
Get the latest gadgets delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
The Information reported that DeepSeek and Moonshot AI relied heavily on outputs from OpenAI’s GPT-6 Astra to train recent models, citing people with direct knowledge. Reuters repeated the report, but the companies had not confirmed or denied it, and public reporting does not disclose the volume or method of data collection.
The Information reported October 28 that Chinese AI labs DeepSeek and Moonshot AI relied heavily on outputs from OpenAI’s GPT-6 Astra to train recently released models. Reuters repeated the claim, naming Moonshot’s Kimi K3.1 Pro and DeepSeek’s V4.1 Flash update, but none of the three companies had confirmed or denied it in Reuters’ account.
The Information’s report, by Jing Yang, Juro Osawa and Qianer Liu, cited people with direct knowledge. Its public opening described Astra as OpenAI’s most powerful released model and said Chinese AI labs were among a growing group of customers seeking to extract its capabilities. The remainder of the article is paywalled, limiting what can be checked from the public record.
Reuters summarized the central allegation as DeepSeek and Moonshot having relied heavily on Astra outputs to train their latest models. The models identified were Kimi K3.1 Pro, released October 23, and DeepSeek-V4.1-Flash-1017, launched October 17. Reuters said OpenAI, DeepSeek and Moonshot did not immediately respond to requests for comment.
The source article also cited South Korean broadcaster SBS for additional details: that labs had previously focused on Anthropic’s Claude and shifted toward OpenAI after Anthropic strengthened defenses. SBS attributed a response to an OpenAI spokesperson saying the company takes the issue seriously and continues to work on it. Those details could not be independently checked against The Information’s paywalled report, so they remain secondhand in the supplied material.
Infographic 01 · Distillation in plain words
How distillation works, and where it crosses a line.
A big model teaches a small one by example. The technique is ordinary; the dispute is about whose model is the teacher and how the answers were obtained.
Send the teacher prompts
A strong “teacher” model gets a huge number of carefully chosen tasks.
Keep answers and reasoning
The step-by-step working is worth far more than the final answer alone.
Teach a student to imitate
A smaller or cheaper “student” model learns to reproduce the teacher’s moves.
Normal practice
The U.S. advisory calls it “a legitimate and useful technique”.
- A lab distills its own large model
- Goal: smaller, faster, cheaper versions
- Open models used within their licence
Where the fight starts
What OpenAI, Anthropic and U.S. agencies allege.
- The teacher is a rival’s model
- Industrial scale: millions of exchanges
- Fake, resold or stolen accounts and keys
- Breaks terms that ban building competing models
Why hidden reasoning is the prize
Anthropic’s own test: a model trained on Claude Opus 4.8 transcripts, scored on SWE-bench Verified (coding).
That is why providers now hide or summarize reasoning, and why extraction attempts target it.
Distillation is a student copying a top tutor’s worked solutions. Legal in a classroom. Contested when the tutor never agreed.Thorsten Meyer · my reading
Sources: NSA/CISA/FBI advisory AA26-251A, 8 Sep 2026; Anthropic, 12 Oct 2026 (vendor’s own experiment). Analogy and framing: my reading.
Powered by Thorsten Meyer AI
Why Model Training Provenance Matters
If the reporting is accurate, it would put a spotlight on how AI labs obtain training material and where ordinary model distillation crosses into a dispute over a rival’s terms of service. Distillation—the use of a larger model’s outputs to train a smaller one—is a standard technique. The allegation is not simply that the labs used distillation, but that they relied substantially on a competing provider’s model.
The commercial stakes are clear in the prices and performance figures cited by ThorstenMeyerAI.com. Its comparison lists Astra at $10 per million input tokens and $50 per million output tokens, compared with $2 and $10 for Kimi K3.1 Pro and $0.30 and $1.20 for DeepSeek’s model at peak hours. On the same independent index, the three models scored 57, 46 and 32, respectively. These figures indicate lower listed prices and lower scores for the two Chinese models on that index; they do not establish that Astra outputs produced those results.
For businesses choosing models, the dispute is a reminder to ask vendors about training-data provenance, applicable terms and fallback options. For users, the source says there is no immediate change to the apps. It advises caution with cheap, unofficial API resellers, though the report does not establish that any specific reseller was involved.
Products Worth Considering
Amazon Kindle Paperwhite Signature Edition (12th Generation, 2024), 32 GB, Metallic Black
Jackery HomePower 3000 Portable Power Station with Two 200W Solar Panels
Amazon Kindle Scribe Colorsoft 64GB (1st Generation)
As an affiliate, we earn on qualifying purchases.
Distillation and Earlier Allegations
In distillation, a larger “teacher” model answers prompts and a smaller “student” model is trained to imitate those answers. AI labs commonly use the method to produce smaller or less costly versions of their own systems. The source material says a U.S. government advisory described distillation as a legitimate and useful research technique. The controversy arises when a lab uses a competitor’s outputs at large scale, particularly if access was obtained through fake or resold accounts or in violation of the provider’s terms.
The report comes amid earlier allegations about access to AI systems. The source article says OpenAI in September linked a “core cluster” of a July effort to extract hidden reasoning from its models to people associated with Moonshot AI. It also says U.S. security agencies named DeepSeek and Moonshot in an advisory, while China’s commerce ministry called the accusations baseless. These are separate claims and responses; they do not independently prove the new allegation about Astra training.
OpenAI’s terms, according to the source, prohibit using its outputs to build competing models. Whether that rule was breached would depend on the conduct and applicable agreements. The source characterizes the potential issue chiefly as contractual, with possible concerns about fraud if access was improperly obtained. That is not a legal finding, and the reported training activity has not been publicly substantiated with methods or records.
Evidence Behind the Astra Allegation
The public material does not specify how many Astra outputs were used, how they were collected, which accounts accessed the model, or how much those outputs contributed to either model’s training. The Information’s article is largely paywalled, and the claim rests on sources described as having direct knowledge rather than publicly presented datasets or records.
There is also no company confirmation or denial in Reuters’ account. Moonshot’s launch post credited Kimi K3.1 Pro’s gains to substantially larger post-training compute and large-scale reinforcement learning; DeepSeek’s update note, according to the source, did not discuss training data. Those statements do not resolve whether Astra outputs were also used. It remains unclear whether the reported conduct violated any contract, involved improperly obtained access, or would have a measurable effect on model performance.
Company Responses and Further Evidence
The next developments to watch are public responses from OpenAI, DeepSeek and Moonshot, along with any additional reporting that supplies details on access, scale or training methods. The supplied source records no confirmed investigation outcome, legal action or technical evidence establishing the allegation.
Organizations evaluating these models can review provider documentation and contract terms, ask vendors to explain training-data provenance, and test candidate systems on their own workloads. The report does not establish a change to the models or apps now in use. Until the companies or additional evidence clarify the matter, the central claim should be treated as a reported allegation rather than a verified account.
Key Questions
What does the report allege?
The Information says DeepSeek and Moonshot relied heavily on GPT-6 Astra outputs to train recent models. Reuters repeated that claim, citing the same report.
Which models are named?
The report identifies Moonshot’s Kimi K3.1 Pro, released October 23, and DeepSeek-V4.1-Flash-1017, launched October 17.
Have the companies confirmed the claim?
No. Reuters reported that OpenAI, DeepSeek and Moonshot did not immediately respond to its requests for comment. The source material provides no confirmation or denial from them.
Is distillation itself prohibited?
Distillation is a common AI training method and was described in a U.S. advisory cited by the source as useful research. The dispute concerns the alleged use of a competitor’s outputs and whether access or use violated OpenAI’s terms; no legal finding is reported.
What evidence has been made public?
The cited public coverage does not provide the volume of outputs, collection method or account details. The central report cites people with direct knowledge, while the full article is paywalled.
Source: ThorstenMeyerAI.com
