[2/7] Before 2025, China barely existed in Anthropic’s writing. The DeepSeek moment drastically increased China mentions, and 2026 Q2 and Q3 also saw an insane surge in China mentions. The last bar is low because we are just one month into 2026 Q4. 2025: 62 mentions. 2026 so far: 356.

MofCom said they will have a near auto licensing regime for open-source AI that allows more rapid release it seems? "To support the development of an open-source AI innovation ecosystem, automatic licensing and other facilitation measures will apply to technology exports involving open-source releases of AI technologies, Chinese companies building and operating their own computing centers overseas, overseas chip tape-outs and fabrication, joint R&D using overseas resources, and participation in international standard-setting—essentially allowing exports to proceed immediately upon filing."

Unitree CEO Wang Xingxing recently said that GPT-7 could solve the challenges of embodied AI entirely and that he had made a mistake by not taking AI seriously enough. Quote: “We’ve put it (GPT-6) on various robots and tested it across a range of tasks, and the results have actually been very good. Our company and other robotics companies are already running pilot deployments in factories. To achieve large-scale adoption, though, both generalization and efficiency still need to improve further. But I think these challenges could very well be resolved over the next few years. My faith in AI grows stronger every year. The biggest mistake I’ve made over the past decade or so has been not believing in AI enough. Deployment in commercial services may happen even faster than in industrial settings.”

South Review, a state media outlet, did full coverage from Coxon's resignation to pacing the frontier to RSI to METR's report on OAI-HF incidents, under the title "'Even Pulling the Plug Won’t Work': AI Scares Engineers Away" I think the last two paragraphs are a great example of how the whole AI safety call may be read from the outside -- with suspicion and mistrust, but not full denial of potential AI risks "Despite all the lofty rhetoric, the debate ultimately comes down to a struggle over the pursuit and distribution of power and profit. A “great slowdown” may sound appealing, but it could amount to false advertising, allowing leading companies to entrench their dominance in the name of safety. Meanwhile, opponents of slowing down invoke innovation and national competitiveness while turning a blind eye to the instances of "AI crossing established boundaries" that have already occurred. The fierce clash of values is, in the end, little more than a struggle over interests. In this race, everyone talks about humanity’s fate, yet perhaps no one truly represents humanity’s interests."

Crazy if true, and it will probably have a significant influence on open-weight model policy in China. Just to summarize because grok English translation for me is not super consistent with the Chinese text: - A “jailbreak model station” is described as an API service that gives users access to models whose safety restrictions have been weakened or bypassed through jailbreak prompts, altered system prompts, or similar methods. - According to screenshots circulating in chat groups, someone allegedly used one of these models in a cyberattack on a .gov.cn website (Chinese government website). - The screenshots reportedly show access to a CMS super-admin account, user records, news content, database backups, and attempts to create persistent access and retrieve configs or keys. - After the incident, the operator of the transfer station reportedly said the service was shutting down, mentioning police involvement, and the associated group was soon dissolved. - Someone in the group also said they had reported the activity to China’s cyber police.