Chinese labs now mainly distill OpenAI's GPT-6 Astra after Anthropic tightens defenses

Chinese AI labs are now mainly tapping OpenAI's GPT-6 Astra instead of Anthropic's Claude to distill model outputs, according to The Information, as reported by The Decoder on Oct 29, 2026.

What was reported

The Information reports that DeepSeek and Moonshot AI leaned heavily on Astra's outputs to train their latest models. People familiar with the matter say this includes Kimi K3.1 Pro and V4.1-Flash 1017.

Moonshot declined to comment. DeepSeek did not respond to a request for comment.

How the distillation works

This kind of distillation only works if the labs get the models to reveal their hidden reasoning steps. OpenAI and Anthropic both ban that practice.

Anthropic's defenses

Claude used to be the go-to target, but Anthropic has tightened its defenses. The company says its new methods cut the amount of data attackers can pull by 90 to 99 percent. That figure is Anthropic's own claim.

OpenAI's position

In late September, OpenAI reported a coordinated distillation campaign that it says involved people linked to Moonshot. An OpenAI spokesperson said the company takes the problem very seriously. Until now, attackers adjusted their tactics every time OpenAI rolled out a new defense.

What this means

The reporting ties the shift in targets to defensive changes at Anthropic. Anthropic says its methods sharply reduce extractable data, and Astra is now reported as the main target. The claims about which models DeepSeek and Moonshot trained on Astra outputs rest on people familiar with the matter. Neither company has commented on them.

Both OpenAI and Anthropic ban extracting hidden reasoning steps, and distillation depends on getting a model to expose them. OpenAI's statement that attackers adapted to each earlier defense is the context for how its own protections are being tested. Anthropic's 90 to 99 percent reduction is a company-reported figure.