🔍 Read the full analysis: GPT-6 Astra And China’s Latest AI Models: The Report’s Findings And Caveats on ThorstenMeyerAI.com
Get business pricing on office and shipping supplies
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
The Information reported that DeepSeek and Moonshot AI relied heavily on GPT-6 Astra outputs when developing recent models, citing people with direct knowledge. Reuters said the three companies did not respond to its requests for comment, and public details about the alleged training, scale and access methods remain unavailable.
The Information reported October 28 that Chinese AI labs DeepSeek and Moonshot AI relied heavily on answers from OpenAI’s GPT-6 Astra to train recent models, citing people with direct knowledge. Reuters repeated the allegation, but said OpenAI, DeepSeek and Moonshot did not immediately respond to requests for comment; publicly available details about the alleged training remain limited.
The models named in coverage are Moonshot’s Kimi K3.1 Pro, released October 23, and DeepSeek-V4.1-Flash-1017, released October 17. Reuters summarized The Information’s central claim as the labs having relied heavily on Astra outputs. The original report’s publicly visible opening describes Chinese AI labs as a growing customer base for Astra, but the rest of the article is paywalled, limiting independent review of its sourcing and supporting evidence.
The public record supplied with the report contains no figures for the volume of outputs, no description of how prompts or model responses were collected, and no explanation of what “relied heavily” means in practice. The labs have not publicly confirmed the allegation in the cited coverage. Moonshot’s launch post attributed Kimi K3.1 Pro’s gains to more post-training compute and large-scale reinforcement learning; DeepSeek’s update note did not describe its training data.
A further account appeared in South Korean broadcaster SBS’s October 29 summary: it said sources told The Information that Chinese labs had focused on Anthropic’s Claude before moving toward OpenAI models after Anthropic tightened defenses. SBS also reported a statement from an OpenAI spokesperson saying the company takes the issue seriously and continues working on it. Those details could not be checked against The Information’s full paywalled report, so they should be treated as secondhand reporting, not independently established facts.
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
How Model Distillation Raises Stakes
The allegation matters because distillation can transfer useful capabilities from a costly model into a cheaper one. If a competitor uses large numbers of another provider’s answers to train its own system, it could reduce the time and expense required to reproduce some capabilities. The report, however, does not establish how much Astra contributed to either model or whether their performance came from that source.
The source material compares listed prices and scores on one independent index: Astra at $10 per million input tokens and $50 per million output tokens, Kimi K3.1 Pro at $2 and $10, and DeepSeek’s model at $0.30 and $1.20 during peak hours. Their reported index scores were 57, 46 and 32, respectively. These figures indicate different prices and scores on that index; they do not prove the models are equivalent, or that any price or performance difference resulted from distillation.
For businesses choosing models, the unresolved provenance question adds a reason to ask providers about training-data sources and access controls, test systems on their own workloads and maintain alternatives. The source article says nothing changes for users in the apps today and advises avoiding cheap, unofficial API resellers. That is practical guidance, not evidence that a particular reseller or user is connected to the allegation.
Products Worth Considering
HP Sprocket Photo Booth Machine Instant Color Photo Printer with 10.1 Touchscreen, LED Ring Light, 250+ Frames, 3×4 ZINK Paper, QR Code Shar
Samsung 27-Inch Odyssey QD-OLED G8 4K 240Hz Gaming Monitor
FLEXISPOT 42-Inch Electric Standing Desk Converter
As an affiliate, we earn on qualifying purchases.
Distillation and Earlier Allegations
Distillation is a standard training method: a smaller “student” model is trained on responses generated by a larger “teacher” model. Labs commonly use the method to make their own systems cheaper or more efficient. A U.S. government advisory cited in the source material describes distillation as a legitimate and useful AI research technique. The disputed issue is not the technique alone, but whether one company used another provider’s outputs at scale and in ways that violated access rules or contract terms.
OpenAI’s terms, as described in the source article, prohibit using its outputs to build competing models. If the reported conduct occurred, that could raise a contract question; questions about fake or resold accounts could also concern how access was obtained. The available reporting does not establish that such accounts were used or determine whether any law was broken. Any legal conclusion would depend on verified facts and applicable terms.
The report follows earlier public claims about efforts to extract model capabilities. The source article says OpenAI linked a “core cluster” of a July campaign targeting hidden reasoning to people associated with Moonshot AI, and that U.S. security agencies named DeepSeek and Moonshot in an advisory. It also says China’s commerce ministry called the accusations baseless. Those are claims and responses in the wider dispute, not independent proof of the specific Astra training allegation.
“DeepSeek and Moonshot “relied heavily on outputs” from OpenAI’s GPT-6 Astra to train their latest models.”
— Reuters’ summary of The Information report
Evidence Behind the Astra Claim
The central allegation remains unconfirmed publicly in the cited reporting. The Information relied on people described as having direct knowledge, but the accessible account does not identify them or provide the underlying records. Reuters said the three companies did not immediately respond; that is not the same as a denial, and no response should be inferred from silence.
It remains unclear how many Astra responses were used, whether the labs obtained them directly or through third parties, which parts of either model were affected, and how much any such training contributed to performance. The report does not provide a method for independently testing the claim. It also does not establish that the labs violated a contract or law. Those questions require evidence not included in the public material summarized here.
What Responses Could Clarify
The next useful developments would be direct responses from OpenAI, DeepSeek and Moonshot, along with any fuller account of the evidence, methods and scale behind The Information’s reporting. Further reporting could also clarify whether access-control records, public statements or other documentation support the allegation. Until then, readers should distinguish the reported claim from established facts about how the models were trained.
Companies evaluating AI providers can ask for information about model provenance and acceptable-use policies, run their own task-specific tests, and keep a fallback system where reliability matters. The source material does not report any change to the models or apps as a result of the story. Any investigation, contractual action or product response remains unannounced in the cited coverage.
Key Questions
What did The Information report?
It reported that DeepSeek and Moonshot AI relied heavily on GPT-6 Astra outputs to train recent models, citing people with direct knowledge. Reuters repeated the claim; the companies had not publicly confirmed it in the cited coverage.
Which models were named?
The report named Moonshot’s Kimi K3.1 Pro, released October 23, and DeepSeek-V4.1-Flash-1017, released October 17.
Is distillation itself improper?
No. Distillation is a common training technique. The dispute concerns whether one lab used a rival’s outputs at scale and whether access or use breached the provider’s terms. The public material does not establish that this happened or that a law was broken.
Have the companies confirmed the report?
The cited Reuters account said OpenAI, DeepSeek and Moonshot did not immediately respond to requests for comment. The source material does not include a direct confirmation from any of them.
Does the report mean users need to change apps?
The source material reports no immediate change to the apps. The allegation does not establish a user-facing service change; businesses may want to review model provenance and test providers against their own requirements.
Source: ThorstenMeyerAI.com
