A new report released Thursday by Anthropic alleges persistent distillation attacks by China-based AI companies, which have escalated in recent months as competition in the space has intensified.

It is one of the most divisive topics in the AI sphere right now, with far-reaching geopolitical implications as a grave add-on: did Moonshot distill its Kimi K3 model from Anthropic's model? Most detractors point to the ~2 week window between the public rollout of Fable and Kimi K3 to suggest that an expansive distillation is virtually impossible here. But, a newly discovered API vulnerability might just be the smoking gun US-based AI labs have been looking for. Moonshot's Kimi K3 demonstrated an unusually high statistical probability of predicting and continuing Claude's text versus other open-weight models like DeepSeek-V4-Flash AI […]

Read full article at https://wccftech.com/kimi-k3-shows-an-uncanny-affinity-to-the-reasoning-patterns-of-claude-opus-as-a-new-study-hands-the-trump-administration-a-smoking-gun/

After Moonshot's Kimi K3 spurred multi-vector allegations of model distillation, China's AI labs are apparently becoming a lot more sensitive to this issue, with TikTok owner ByteDance becoming the second such AI lab to explicitly ban distillation from US models. Meanwhile, Anthropic has just confirmed that it's working on a dedicated ASIC for Claude, potentially extending its inference-related economies of scale. China's ByteDance commits it won't resort to distillation as a shortcut for advancing its AI models For the benefit of those who might not be aware, Moonshot's Kimi K3 model jumped right into the ongoing multi-dimensional struggle between China […]

Read full article at https://wccftech.com/chinas-bytedance-bans-employees-from-distilling-us-ai-models-as-anthropic-confirms-its-building-a-custom-chip-for-claude/

In a dramatic development that should surprise no one in hindsight, a high-level official in the Trump administration has just declared that Moonshot's Kimi K3 model has been distilled from Anthropic's Fable, and further trained on NVIDIA's GB300 GPUs in Thailand. Michael Kratsios asserts that Moonshot covertly owns some NVIDIA GB300 servers, and accessed additional GB300 GPUs in Thailand to train its Kimi K3 model As we reported recently, the proclivity of China's AI labs to distill frontier US models is raising hackles within the Trump administration, especially given the fact that there is no easy way of preventing this. […]

Read full article at https://wccftech.com/moonshot-owns-secret-nvidia-gb300-servers-and-tapped-thailands-gb300-stash-to-distill-kimi-k3-from-anthropics-fable-as-per-solid-evidence-from-a-trump-administration-official/

Moonshot's Kimi K3 AI model has suddenly taken the world by storm, jumping to the apex spot in Arena's front-end code rankings seemingly out of the blue. Yet, below the surface, a growing quantum of anecdotal evidence indicates that the Kimi K3 has been distilled from Anthropic's Claude, with the model going so far as to identify itself as such in at least one conversation that is doing the rounds on social media. The Kimi K3 model's training and inference cost efficiency is poles apart, which indicates an underlying discrepancy As we explained in a recent post, Moonshot has just […]

Read full article at https://wccftech.com/chinas-kimi-k3-identifies-itself-as-anthropics-claude-in-at-least-one-conversation-betraying-its-distilled-origins/

"Distillation" is a hot topic as frontier labs try to prevent smaller competitors from copying their models.

Anthropic accuses DeepSeek, Moonshot, and MiniMax of using 24,000 fake accounts to distill Claude’s AI capabilities, as U.S. officials debate export controls aimed at slowing China’s AI progress.