Context

Axios reports that (1) the US government requested that Anthropic delay the release of Mythos

The administration tried to get Anthropic to pause releasing the latest models but was unsuccessful, the official said, prompting the export control letter.

and (2) the US government then put export controls on the model

The Trump administration is blocking foreign governments, companies and individuals from accessing Anthropic's most advanced AI models.

This forced Anthropic to withdraw the model release (announcement). It is no longer widely available.

Resolution

The spirit of this market is to try and capture whether something comparable will happen to an OpenAI model by EOY. This definition will not be particularly legalistic, just whether it seems sufficiently comparable. For the purposes of this question, if Anthropic had delayed the release of the model after the request in (1), I would consider that to be an example of "delay" (due to the US govt request) and that would be sufficient for YES. And Anthropic's choice to suspend access to Mythos absolutely counts as a "withdraw" and would also resolve it YES.

This will resolve based on a consensus of credible media reporting (particularly for (1), given that the reasons won't be public). To count as "withdraw", it need not be complete—what matters is functional access to the model (roughly—a large fraction of those who could previously access the model before no longer can). In each case, what matters is if this happens under pressure from the US government. If Anthropic had such solid KYC in place that it complied with the US govt order by simply suspending access to non-Americans, that would count for YES (although this question is strictly about OpenAI models).