🔍 Read the full analysis: Could China’s New AI Models Have Learned From GPT-6 Astra? on ThorstenMeyerAI.com
Get business pricing on tech for your team
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
The Information reported on Oct. 28 that DeepSeek and Moonshot AI relied heavily on outputs from OpenAI’s GPT-6 Astra to train recently released models. Reuters said the three companies did not respond to its requests for comment; the public reporting does not disclose the scale or method, and the claim remains unconfirmed.
The Information reported Oct. 28 that Chinese AI labs DeepSeek and Moonshot AI relied heavily on answers from OpenAI’s GPT-6 Astra to train recently released models, a claim that could raise questions about the use of rival systems’ outputs and OpenAI’s service terms. Reuters summarized the report, but said none of the three companies had responded to its requests for comment; the public evidence does not establish how the outputs were obtained or how much they influenced either model.
The models named in the reporting are Moonshot’s Kimi K3.1 Pro, released Oct. 23, and DeepSeek-V4.1-Flash-1017, released Oct. 17. The Information’s Oct. 28 story, by Jing Yang, Juro Osawa and Qianer Liu, cited people with direct knowledge. Its publicly visible portion described Chinese AI labs as a growing customer base for Astra; the detailed account is behind a paywall. Reuters reported the central allegation that the labs relied heavily on Astra outputs.
The public record provided so far does not include data volumes, prompts, account details or training records that would independently show the claimed use. Moonshot’s launch post attributed Kimi K3.1 Pro’s gains to substantially more post-training compute and large-scale reinforcement learning. DeepSeek’s update note, according to the source material, did not describe its training data. Those statements do not resolve whether Astra outputs were also used.
The report’s publication is not proof of a breach or of unlawful conduct. OpenAI’s terms prohibit using its outputs to build competing models, according to the source material. Whether the reported conduct occurred, violated those terms or involved improper access has not been established publicly. Distillation is a standard training technique; the allegation concerns whether one company’s model outputs were used at scale to develop competitors.
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-Output Access Matters
If substantiated, the allegation would put a spotlight on how AI developers can learn from a rival’s system through repeated access to its responses. A model’s outputs can convey useful capabilities without access to its underlying weights, but the degree of learning depends on the data, training method and volume. Those details are not disclosed in the public reporting.
The commercial stakes include both training costs and access rules. The source material 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 also gives scores of 57, 46 and 32, respectively, on an unnamed independent index. These figures suggest differences in price and performance on that index; they do not show that any model was derived from another.
For businesses choosing AI tools, the immediate practical point is not to treat the allegation as proof that the named models are unreliable. It is a reason to ask vendors about training-data provenance, contractual terms and evaluation methods, and to test models on the company’s own work. The source material also advises keeping an alternative model available and avoiding unofficial API resellers. It reports no immediate change to the consumer apps.
Products Worth Considering
Bose QuietComfort Wireless Bluetooth Headphones with Active Noise Cancellation, Mic, USB‑C Charging, Deep Bass, 24‑Hour Battery – White Smok
Samsung 27 Odyssey OLED G5 (G50SF) QHD QD-OLED Gaming Monitor
Apple iPad Air (2025) 13‑inch Wi‑Fi 256 GB – Space Gray (M3 chip)
As an affiliate, we earn on qualifying purchases.
Earlier Disputes Over Distillation
Distillation involves training a smaller model on responses generated by a larger “teacher” model, so the smaller system learns to produce similar outputs. Developers commonly use the method with their own models. A U.S. government advisory described distillation as “a legitimate and useful technique in AI research,” according to the source material. The controversy in this case is whether a competitor’s outputs were used at industrial scale and whether access complied with the provider’s rules.
The allegation follows earlier concerns about attempts to extract capabilities from AI services. The supplied account says OpenAI linked a “core cluster” of a July campaign targeting hidden reasoning to individuals associated with Moonshot AI, and that U.S. security agencies named both DeepSeek and Moonshot in an advisory. It also says China’s commerce ministry called the accusations baseless. These are separate claims and responses; they do not independently verify The Information’s account about the October model releases.
A further detail attributed to South Korea’s SBS on Oct. 29 was that the labs had previously focused on Anthropic’s Claude and shifted attention to OpenAI after Anthropic tightened defenses. SBS also quoted an OpenAI spokesperson as saying the company takes the issue seriously. The original report’s full detail was not publicly available in the material supplied, so those points should be treated as secondhand reporting, not independently confirmed findings.
Evidence Behind the Allegation
The central claim remains unconfirmed by the named companies in the public material. Reuters reported that OpenAI, DeepSeek and Moonshot did not immediately respond. The Information cited sources with direct knowledge, but its detailed reporting is paywalled, and the supplied public account does not provide documents, datasets, exact output volumes or a technical account of how Astra responses entered training.
It is also unclear whether the alleged use, if it occurred, involved ordinary API access, accounts obtained through intermediaries, or another method. Claims about fake or resold accounts appear in the source’s explanation of the broader dispute, not as established details of this specific case. No public finding here establishes a contract violation, fraud or other legal breach. The companies’ own explanations of their models’ training do not settle the allegation either.
Company Responses and Further Evidence
The next useful developments would be responses from OpenAI, DeepSeek and Moonshot, along with any additional reporting that specifies the volume and source of the training outputs. Evidence such as account records, training documentation or a detailed technical analysis could help assess whether the reported use occurred and whether it materially shaped the released models.
Until then, readers and customers should distinguish the report’s allegation from confirmed facts. Companies evaluating the models can review vendor disclosures and conduct their own task-specific tests; users should rely on official services rather than unverified API resellers. No immediate product change is reported in the supplied material, and the legal or contractual status remains unresolved.
Key Questions
What did The Information report?
It reported that DeepSeek and Moonshot AI relied heavily on GPT-6 Astra outputs to train models released this month: DeepSeek-V4.1-Flash-1017 and Moonshot’s Kimi K3.1 Pro. The report cited people with direct knowledge.
Have the companies confirmed the claim?
No confirmation is included in the public material supplied. Reuters said OpenAI, DeepSeek and Moonshot did not immediately respond to its requests for comment.
Is distillation itself improper?
Not necessarily. Distillation is a standard technique for training a model using another model’s outputs, and a U.S. government advisory described it as a legitimate research technique. The dispute is about the alleged use of a competitor’s outputs at scale and whether access followed the provider’s terms.
Does the report prove the models copied Astra?
No. The public account does not provide the training records, output volumes or method needed to establish how Astra outputs were used or how much they affected the models. The allegation remains unverified publicly.
What should AI customers do now?
The source material reports no immediate change to the apps. Businesses can ask vendors about data provenance and terms, test models on their own tasks and maintain alternatives. Users should avoid unofficial API resellers.
Source: ThorstenMeyerAI.com
