Some three dozen American technology firms have signed a joint statement urging Washington to back open-weight artificial intelligence models—those anyone can download, inspect, modify, and run on their own hardware—as the foundation of what they call American AI leadership, published July 24 by Microsoft. In response to Kimi K3, some had called on the U.S. government to ban the use of such open-weight models.
The signatories span the industry: Microsoft, OpenAI, Meta, NVIDIA, IBM, Cisco, Dell, Mistral, Hugging Face, Mozilla, the Linux Foundation, Y Combinator, Andreessen Horowitz—and Palantir, whose AI systems the Pentagon uses for targeting. Absent (at least at this time) are Anthropic and Google DeepMind.
Their argument draws the analogy to open-source software, which “did more than lower the cost of software,” the statement says. It formed a foundation of collaborative development upon which engineers have built for decades. Open weights, they argue, would widen access for startups, universities and public institutions, prevent lock-in to a single provider, and—counterintuitively—improve safety, since concentrating advanced capability behind a few closed models creates “a small number of single points of failure,” while open models let many teams find and fix flaws. They call for expanding compute access for researchers, investing in shared datasets and evaluation tools, and avoiding “premature restrictions” that would drive innovation overseas.
