🔍 Read the full analysis: How Much Does The Report Tell Us About China’s AI Models And GPT-6 Astra? 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 on October 28 that DeepSeek and Moonshot AI relied heavily on OpenAI’s GPT-6 Astra outputs to train recent models. Reuters reported that none of the three companies had immediately responded; public details about the scale, methods and evidence remain limited.
The Information reported October 28 that Chinese AI companies DeepSeek and Moonshot AI relied heavily on outputs from OpenAI’s GPT-6 Astra while training recent models, including DeepSeek’s V4.1 Flash update and Moonshot’s Kimi K3.1 Pro. Reuters summarized the report within hours, but said OpenAI, DeepSeek and Moonshot had not immediately responded, leaving the allegation unconfirmed by the companies.
The Information’s report, by Jing Yang, Juro Osawa and Qianer Liu, cited people with direct knowledge, according to the account provided by ThorstenMeyerAI.com. Its publicly visible introduction described Chinese AI labs as a growing customer group seeking to extract Astra’s capabilities; the rest of the article is paywalled. Reuters reported the central claim as heavy reliance on Astra outputs to train the companies’ latest models.
The models named were Moonshot’s Kimi K3.1 Pro, released October 23, and DeepSeek-V4.1-Flash-1017, launched October 17. The report said both benefited from distillation using Astra outputs. It did not provide public figures for how many outputs were used, describe the training method, or explain how access to those outputs was obtained.
The models’ own release material offers different explanations of their development, but does not settle the allegation. Moonshot credited Kimi K3.1 Pro’s gains to substantially larger post-training compute and large-scale reinforcement learning. DeepSeek’s update note, as described by the source, did not discuss training data. Neither statement confirms nor rules out the use of another company’s model outputs.
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 report is accurate, the dispute concerns whether a company can use a rival’s outputs at scale to develop a competing model, and whether those outputs were obtained in ways that violated the provider’s terms. It could affect model-provider contracts, access controls and the information businesses seek when evaluating the origin of AI systems. The public evidence described so far does not establish that any law or contract was breached.
The source compares listed prices and scores to show why the allegation has drawn attention. It gives Astra’s price as $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, it reports scores of 57, 46 and 32, respectively. These figures show lower listed prices and lower scores for the two named models on that index; they do not prove how either model was trained or establish performance across all tasks.
For businesses, the practical response is to ask vendors about data and model provenance, test systems on their own workloads and retain alternatives. For users, the source says there is no reported change to the named apps’ availability. It advises caution around cheap, unofficial API resellers, though the report itself does not establish that a particular reseller supplied access in this case.
Products Worth Considering
Bose QuietComfort Wireless Bluetooth Headphones with Active Noise Cancellation, Mic, USB‑C Charging, Deep Bass, 24‑Hour Battery – White Smok
Logitech MX Mechanical Wireless Illuminated Keyboard Tactile (Graphite)
Jackery HomePower 3000 Portable Power Station with Two 200W Solar Panels
As an affiliate, we earn on qualifying purchases.
Distillation and Earlier Allegations
Distillation is a standard model-training technique: a smaller student model learns from answers generated by a larger teacher model. AI labs commonly use it to produce less costly or faster versions of their own systems. The technique itself is not evidence of misconduct. The issue raised by this report is whether rival-model outputs were used at scale and whether access or use conflicted with a provider’s terms.
The source says OpenAI’s terms prohibit using its outputs to build competing models. Whether that restriction applies to particular conduct, and whether any conduct occurred, would depend on evidence and legal interpretation. The source frames the possible dispute chiefly as a contract question, with a separate possibility of fraud if access was obtained through deceptive accounts. That is analysis, not a finding by a court or regulator.
There is earlier public reporting about attempts to obtain model outputs. The source says OpenAI linked a July campaign targeting its hidden reasoning to individuals associated with Moonshot AI, and that U.S. security agencies named both DeepSeek and Moonshot in an advisory. China’s commerce ministry has called accusations against Chinese AI companies baseless. These are distinct claims and responses; they do not independently verify the new allegation about training Kimi K3.1 Pro or DeepSeek’s October update.
“Distillation is “a legitimate and useful technique in AI research.””
— The U.S. government advisory, as described by ThorstenMeyerAI.com
Evidence Behind the Distillation Claim
The cited coverage does not disclose the volume of Astra outputs allegedly used, the prompts or training procedure involved, or whether the outputs came through direct subscriptions, resellers or another route. The Information cited people with direct knowledge, but the public account supplied here does not identify them or provide documentary evidence. Reuters said the three companies had not immediately responded; that is not the same as a definitive denial or confirmation.
A separate report by South Korea’s SBS said the labs had previously focused mainly on Anthropic’s Claude and shifted toward OpenAI after Anthropic tightened defenses. The source says it could not verify those details against The Information’s full article. SBS also attributed a statement to an OpenAI spokesperson saying the company takes the issue very seriously and continues to work on it. These points should be treated as separately attributed reporting, not as independently established facts about the training of the named models.
It is also unclear whether any investigation is underway, whether OpenAI has presented evidence to the companies or authorities, or whether the dispute will result in enforcement or litigation. The material provided contains no independent technical analysis establishing that Astra outputs were present in either model’s training data.
Responses and Evidence to Watch
The immediate next development would be a response from OpenAI, DeepSeek or Moonshot addressing the specific claim. Further reporting could clarify the sources’ basis, the alleged scale of output use and how access was obtained. Any company statement should be distinguished from independently verifiable evidence, such as disclosed records or a technical analysis.
For now, the report is a serious but unverified allegation, not a public finding of misconduct. Readers should watch for a fuller account beyond the paywalled report, responses from the companies, and any statements from regulators or courts. Until those emerge, the effect on the named models’ products and users is not established.
Key Questions
What does the report allege?
The Information reported that DeepSeek and Moonshot AI relied heavily on GPT-6 Astra outputs to train recent models, including DeepSeek-V4.1-Flash-1017 and Kimi K3.1 Pro. The companies have not confirmed the claim in the cited coverage.
What is distillation in AI?
Distillation is a technique in which a smaller model is trained using answers from a larger model. Labs often apply it to their own systems. The dispute in this report concerns alleged use of a rival’s outputs and the scale and terms of that use.
Has OpenAI confirmed that the models used Astra?
No confirmation is included in the cited reporting. Reuters said OpenAI, DeepSeek and Moonshot had not immediately responded to its request for comment. The public details about the allegation remain limited.
Does the report prove that a law was broken?
No. The report, as summarized here, does not establish a legal violation. OpenAI’s terms are described as barring use of its outputs to build competing models, but whether a breach occurred would require evidence and legal assessment.
What should businesses and users do now?
Businesses can ask AI vendors about model and data provenance, test systems on their own tasks and keep alternative providers available. The source reports no immediate change for users of the named apps and advises caution with cheap, unofficial API resellers.
Source: ThorstenMeyerAI.com
