Per Solidot, dozens of tech giants including NVIDIA, Microsoft, Cisco, Dell, IBM, and even OpenAI have co-signed an open letter calling on the US government not to prematurely restrict open-weight models, arguing that doing so would throttle competition or slow domestic innovation. The letter points out that open-weight models foster competition, ensure AI's benefits are broadly shared, and keep the technology out of the hands of a few; closed-source models don't mean absolute safety. Notably absent from the signature list are Google, Anthropic, and xAI/SpaceX — Anthropic is pouring $40 million into lobbying the Trump administration to tighten AI model oversight, while Musk has voiced verbal support for the letter on social media. This line-up takes the open-weights route debate straight to Washington's regulatory table: on one side, an industry alliance of NVIDIA + Microsoft + OpenAI; on the other, an "AI safety first" camp led by Anthropic. In the short term, the regulatory framework will likely remain guiding rather than binding; in the long run, the policy gap between the US and China on open weights will directly determine the real-world share of the next generation of open-source flagships (Kimi K3, Qwen3.8, Mistral 3, Llama 4) among global developers. Whoever can extend the engineering dividend of the open-source ecosystem without crossing national-security red lines will have captured the channel-pricing power in AI before 2027. On the technical side, the most interesting side-effect of all this is that the leading players continue to roll "effort level", model routing, and prompt cache into closed-source engineering weapons, while ceding training data and weight distribution — the "upstream discourse power" — to the open-source side. Open weights isn't charity; it's a chip in the geopolitical game.