Unconfirmed: a post from @theinformation says Anthropic's enforcement against model distillation has pushed Chinese AI labs toward OpenAI's "GPT-6 Astra," but Automatica has not confirmed the claim.
@theinformation wrote on 30 October that, "Exclusive: Anthropic’s crackdown on distillation has pushed Chinese AI labs toward OpenAI’s GPT-6 Astra, now a major source of training data for DeepSeek and Moonshot" @theinformation on X1. That is the full public evidence provided here for the allegation. The account clears Automatica's threshold for reporting an unfamiliar X source on follower count, but those metrics do not verify the underlying claim.
No corroborating company statement, filing, product page or other primary documentation in the available evidence shows an Anthropic policy change, a distillation enforcement action, or a resulting switch by DeepSeek or Moonshot to OpenAI training inputs. Anthropic, OpenAI, DeepSeek and Moonshot have said nothing in the materials provided. The post links to a fuller story, but that text is not part of the evidence available for this report, so the article can verify only that @theinformation made the claim, not that it is correct.
One point that can be checked against existing reporting is the model name. Automatica previously reported on an OpenAI release called GPT-6.1 Terra, with launch timing and API pricing details published on 28 October Automatica Press5. That does not disprove @theinformation's reference to GPT-6 Astra, but it does underscore the sourcing gap: the available evidence here includes no primary product information for any OpenAI model by that name.
What would confirm the rumour is straightforward: a public statement from Anthropic describing a distillation enforcement action; a statement from OpenAI, DeepSeek or Moonshot acknowledging the alleged training-data shift; or primary records showing GPT-6 Astra being used as training data by the named labs. What would kill it would be a denial from the companies involved or documentation showing that no such shift occurred.
As it stands, the numbers are thin. There is one eligible social post and no independent corroboration in the available evidence. That is enough to report that talk is circulating, but not enough to establish that Anthropic changed enforcement in a way that redirected Chinese labs to an OpenAI model.