Somewhere in the steady churn of a field racing to build ever more capable machines, a single line has surfaced that asks an old question in a new form: who learns from whom, and by what right. A report, relayed through Reuters from the reporting of The Information, alleges that two Chinese AI companies, DeepSeek and Moonshot, used OpenAI's GPT-6 Astra in training their newest models. The claim arrives thin, with its method, its evidence, and any answer from the parties all still unseen, so it stands as an allegation to be weighed rather than a settled fact.
The Information Reports DeepSeek and Moonshot Trained Latest Models Using OpenAI's GPT-6 Astra
The whole story arrives as a single line, with no body beneath it.
Mark
So what actually happened here, as far as we know?
Mimi
A report from The Information, relayed by Reuters, says DeepSeek and Moonshot used OpenAI's GPT-6 Astra to train their latest models.
Mark
Why does that matter?
Mimi
If true, it would mean one company's model helped build competitors' models, which raises questions about terms of service and intellectual property.
Luke
Careful with "if true." What we have is a headline. Nothing in it tells us how the models were allegedly used.
Mark
Is there any evidence cited?
Luke
Not in what we were given. We don't know The Information's sources or what documents, if any, back this up.
Mimi
And no one has responded in the material we have. OpenAI, DeepSeek and Moonshot are all silent on the page.
Mark
The keywords mention distillation. Is that what this is?
Luke
That's a label attached to the story, not something the report confirms here. It's a plausible frame, nothing more.
Mark
So what should a reader watch for?
Mimi
Official responses from the three companies, and any enforcement action that follows.
Luke
And the full original report. Until then, this is an attributed claim, secondhand, with real gaps.
El Pulso
- A single headline alleges that DeepSeek and Moonshot trained their newest models using OpenAI's GPT-6 Astra, and the claim is already moving across the industry.
- The story has no body text, no named sources, and no methods, leaving the central accusation hanging on a secondhand report.
- Neither OpenAI, DeepSeek, nor Moonshot has answered publicly, so the silence itself is becoming part of the tension.
- The keywords point toward model distillation, a contested practice, but the mechanism remains unexplained and unconfirmed.
- What comes next depends on the full reporting, company responses, and whether terms-of-service or intellectual property enforcement follows.
Somewhere in the steady churn of a field racing to build ever more capable machines, a single line has surfaced that asks an old question in a new form: who learns from whom, and by what right. A report, relayed through Reuters from the reporting of The Information, alleges that two Chinese AI companies, DeepSeek and Moonshot, used OpenAI's GPT-6 Astra in training their newest models. The claim arrives thin, with its method, its evidence, and any answer from the parties all still unseen, so it stands as an allegation to be weighed rather than a settled fact.
The report arrives as a single line. Reuters, passing along work by The Information, says that DeepSeek and Moonshot, two Chinese artificial intelligence companies, used OpenAI's GPT-6 Astra to train their latest models. That is nearly everything the source offers: a headline, attributed to someone else, with nothing beneath it.
The parties are named with precision, and the allegation is specific. It is not that the systems were merely tested against one another, but that GPT-6 Astra played a part in training the newest models from the two firms. Yet the surrounding details are missing. The report does not explain how the training allegedly happened, what evidence supports it, who spoke to the reporters, or when any of it took place.
No one responds in the material. OpenAI, DeepSeek, and Moonshot offer no statements, and there is no sign they were asked. Reuters appears to be repeating another outlet's reporting without independent confirmation.
The story's keywords point toward model distillation, the practice of training one model on another's outputs, but that is only context for how such claims usually take shape. For now, the allegation stands as reported, awaiting the full account and whatever the companies choose to say.
The report arrives as a single line. Reuters, relaying work by The Information, says that DeepSeek and Moonshot, two Chinese artificial intelligence companies, used OpenAI's GPT-6 Astra to train their latest models. That is the whole of what the source provides: a headline, attributed to another outlet, with no body text beneath it.
The claim is specific in its parties. DeepSeek and Moonshot are named, as is the OpenAI system said to have been involved, GPT-6 Astra. The allegation is that the system was used in training the newest models from the two Chinese firms, not merely tested or benchmarked against them.
Beyond that, the available material says little. It does not describe how the models were allegedly used in training, whether through outputs collected from the system or by some other route. It does not say what evidence The Information relied on, who its sources were, or when the alleged training took place. It does not say how many models are involved or which ones.
Nor does the source carry any response. There is no statement from OpenAI, DeepSeek, or Moonshot, and no indication that any of them were asked or declined to comment. Reuters is passing along another publication's reporting, and the headline shows no sign that Reuters verified the claim independently.
The keywords attached to the story point toward model distillation, the practice of training one model on the outputs of another. That is context for how such an allegation is usually framed, not a confirmed description of what happened here. The source itself does not use the term or explain the mechanism.
What the story points toward is a set of open questions. The first is whether OpenAI, DeepSeek, or Moonshot respond publicly. The second is whether any enforcement follows, whether under terms of service or intellectual property claims. Until detail from The Information's full report or from the companies emerges, the allegation stands as reported, not as established.
Citas Notables
DeepSeek and Moonshot used OpenAI's GPT-6 Astra to train their latest models.— The Information, as relayed by Reuters
Quieres la nota completa?
Lee el original en Reuters ↗
Technology