Nvidia, Microsoft, and Meta have jointly urged the U.S. government not to impose heavy regulation on open-weight AI models, according to CNBC.
Open-weight models make their internal parameters publicly available, letting researchers, startups, and developers build on state-of-the-art systems without the cost of training from scratch — unlike closed models from companies such as OpenAI and Anthropic.
According to CNBC's report, the three companies argue that strict licensing or liability rules aimed at open models would raise compliance costs that only large, well-funded labs with closed systems could absorb, effectively squeezing out startups and academic researchers. They also contend that open access to model weights supports safety research by letting outside groups audit and benchmark systems that would otherwise be opaque.
The intervention comes as U.S. and other governments continue to weigh rules addressing risks from advanced AI, including misuse in cyberattacks and generation of harmful content. Critics of open-weight releases argue that publicly available model weights are harder to control once released, since safeguards can be stripped out by anyone with access to the underlying parameters.
The positions attributed to Nvidia, Microsoft, and Meta reflect a shared interest: all three have invested heavily in open or partially open model ecosystems, from Meta's Llama models to Microsoft's support for open-weight releases on Azure and Nvidia's hardware and software tooling for developers running open models locally.