Anthropic has a trick Unable to withstand being attacked from both front and back. Codex is making a loud noise about Claude's code at the beginning. GLM, DeepSeek, Kimi, and Qwen in China are lagging behind, and the difference in model capabilities is getting smaller and smaller. Anthropic's response was to publish a sour policy white paper calling on the US government to block Chinese AI This move, to put it simply, is to use regulators as a shield. I can summarize the core argument of the article as follows: 1) Computing power is the lifeblood of AI, and the United States has suppressed China through export controls 2) But domestic large models have been following the forefront without falling behind by conducting distillation attacks and massively crawling the outputs of American models for reverse training 3) So we need to plug these two loopholes and lock in the lead for 12-24 months. On the data level, the article cites a figure: in 2026, Huawei's total computing power will only be 4% of Nvidia's, and will further decrease to 2% in 2027. This gap is indeed real, export control played a good card. But what makes me feel uneasy about this article is that Anthropic presents itself as a defender of democratic lighthouses, but the essence of its actions is lobbying policies to help restrict its competitors. A technology company that cannot win in business competition and goes to the government to cut off food for its competitors is a matter of capability anxiety, not strategic confidence. The problem with distillation attacks is real. China's domestically produced large models have reevaluated the boundaries of scaling law in the entire industry with a cost efficiency of less than one tenth. They are open source and can be used, and Anthropic feels the pressure completely reasonable. But the solution is to build one's own moat on policy barriers, and this path has a clear ceiling: policies can change, but technological gaps are the real gap. A truly capable player should use a faster iteration speed to pull the lead, rather than telling Washington in a white paper to help me shut my opponent out. The real audience for this article is not the public, but Congress and the White House. So, this is a beautiful political PR, not a candid declaration of technological competition. I have no doubts about the ability of Anthropic, and it is still the top one in programming. However, my judgment on the signal significance of this action is that it is purely painful.