🔍 Read the full analysis: What The Report Doesn’t Settle About China’s AI Models And GPT-6 Astra 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 to train recently released models, citing people with direct knowledge. Reuters said the three companies did not immediately respond; public evidence does not establish how many outputs were used, how they were obtained, or how much they shaped the models.
The Information reported on October 28 that Chinese AI labs DeepSeek and Moonshot AI relied heavily on OpenAI’s GPT-6 Astra outputs to train recently released models. The claim, attributed to people with direct knowledge, concerns DeepSeek’s V4.1 Flash update and Moonshot’s Kimi K3.1 Pro; Reuters said the three companies did not immediately respond, and the public record does not establish the scale or method of any use.
Reuters summarized The Information’s report as saying the two labs’ latest models “benefitted from distilling Astra.” The named products are DeepSeek-V4.1-Flash-1017, released October 17, and Moonshot’s Kimi K3.1 Pro, released October 23. The Information’s public article opening described Chinese AI labs as a growing customer base for Astra, but the rest of its report is behind a paywall. The detailed allegation therefore rests on reporting that readers cannot fully assess from the public material.
Reuters reported that OpenAI, DeepSeek and Moonshot did not immediately respond to its requests for comment. The source article also cites South Korean broadcaster SBS as reporting that an OpenAI spokesperson said the company takes the issue seriously and continues working on it. That statement is second-hand in the supplied material, and the original Information article could not be independently checked for it.
There are no public figures in the cited coverage for the number of Astra outputs, the prompts used, or the share of training attributable to them. 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 article, did not discuss training data. Those statements do not resolve whether Astra outputs were used.
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, it would put a spotlight on how one AI company’s paid model access may be used by another lab to build a competing system. OpenAI’s terms, as described in the source material, prohibit using its outputs to build competing models. Whether a particular use breached those terms depends on evidence about access, conduct and the applicable agreement; the allegation by itself does not establish a legal violation.
The dispute also matters to customers choosing models on price, capability and reliability. The source article 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. It reports scores of 57, 46 and 32 respectively on the same independent index. Those figures describe prices and results in the cited comparison, not proof that any model copied another or that performance is equal.
For businesses, the report is a reason to ask vendors how training data and model outputs are governed, and to test models on their own tasks rather than rely on headline scores. For users, the source article says there is no immediate change to the apps. It advises caution with cheap, unofficial API resellers, though the supplied material does not link any named reseller to this 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
Dell XPS 13 9350 Laptop, 13.4 FHD+ Display, Intel Ultra 7 256V, 16GB DDR5, 1TB SSD
Apple iPad Air (2025) 13‑inch Wi‑Fi 256 GB – Space Gray (M3 chip)
As an affiliate, we earn on qualifying purchases.
How Distillation Became the Dispute
Distillation is a standard training technique: a student model learns from answers generated by a larger teacher model. Labs commonly apply it to their own systems to create smaller or less costly models. A U.S. government advisory cited by the source calls distillation a “legitimate and useful technique in AI research.” The concern here is not the technique alone, but the allegation that a rival’s outputs were used at scale, potentially through access that violated service terms.
The report arrives amid earlier accusations. The source article says that in September OpenAI linked a “core cluster” of a July campaign seeking 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 attributed claims and responses, not independent confirmation of the current report.
The supplied article also describes earlier allegations involving other labs and says providers have tightened defenses against attempts to obtain hidden reasoning. It notes that SBS reported Chinese labs had previously focused on Anthropic’s Claude before shifting toward OpenAI after Anthropic strengthened protections; the writer says that detail could not be checked against The Information’s full report. These earlier accounts provide context, but they do not verify the new claim about Astra.
“DeepSeek and Moonshot “relied heavily on outputs from OpenAI’s powerful GPT-6 Astra model to train their latest models.””
— The Information, as summarized by Reuters
What the Public Record Cannot Show
The central claim remains unconfirmed by the companies in the supplied coverage. The reporting does not disclose the volume of Astra outputs allegedly used, the prompts or accounts involved, whether access was direct or through intermediaries, or how much those outputs influenced either model’s final training. Without those details, the scale and technical significance of the alleged distillation cannot be independently evaluated.
It is also unclear whether the reported activity violated OpenAI’s terms, how any account access was obtained, or whether the companies dispute the report. Those are separate questions: a model may be trained using outputs without the public record showing how they were collected, and a reported contract restriction is not itself a finding of breach. No regulator or court determination is included in the source material.
The source article’s account of SBS’s additional details is not a substitute for the paywalled reporting or direct statements from the companies. Launch notes that name other training methods also do not settle the issue. More documentation or on-record responses would be needed to establish what happened.
Responses and Evidence to Watch
The immediate next development would be a direct response from OpenAI, DeepSeek or Moonshot addressing whether Astra outputs were used and, if so, under what conditions. The supplied reports do not give a timetable for such responses, a planned investigation, or any announced enforcement action.
Readers should watch for supporting evidence beyond anonymous-source summaries, including company statements, disclosed training methods, account-access records or findings from an independent inquiry. Until then, the report should be treated as an allegation rather than an established account of how either model was trained. The available material does not indicate a change to users’ access to the models.
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. The claim is attributed to people with direct knowledge and has not been confirmed by the companies in the supplied coverage.
Which models are named?
The report names 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 way to train a smaller model using outputs from a larger one. The disputed questions are whether a rival’s outputs were used, how they were obtained, and whether that use complied with the provider’s terms.
Has OpenAI confirmed a breach of its terms?
No confirmation or legal finding is cited in the supplied reports. OpenAI’s terms are described as prohibiting the use of its outputs to build competing models, but the report does not establish the specific conduct or whether a breach occurred.
Does the report change anything for users now?
The source article says there is no immediate change to the apps. It does not report service restrictions or a new user-facing policy tied to the allegation.
Source: ThorstenMeyerAI.com
