Open-weight AI models have drawn public backing from a coalition of prominent U.S. technology executives, including NVIDIA CEO Jensen Huang, OpenAI CEO Sam Altman, Microsoft CEO Satya Nadella, and Meta CEO Mark Zuckerberg. The leaders argue that open-weight AI expands competition, strengthens security, and broadens access to advanced AI tools, urging policymakers to support open-weight models rather than impose broad restrictions on advanced AI. The endorsement comes amid an active policy debate in Washington over how to regulate AI development, including discussions around export controls, reporting requirements for frontier models, and how the United States can maintain competitiveness with China in advanced AI capabilities.

The joint endorsement came after NVIDIA published a letter titled "Open Weights and American AI Leadership," which contends that downloadable and modifiable AI models can strengthen competition, expand access to AI tools, improve cybersecurity, and bolster American innovation. Huang shared the letter in a post on X, while Altman, Nadella, and Zuckerberg each endorsed its message through separate posts.

Industry Leaders Voice Support

Jensen Huang stated that the world needs both frontier-closed and frontier-open models, arguing that open models reinforce safety, cybersecurity, innovation, and national sovereignty. His post on X can be found here.

Sam Altman responded that he wants the United States to lead in both open-source and proprietary AI models and welcomed the initiative. His response is available here.

Mark Zuckerberg described open source as an important force that empowers people and helps prevent centralization. His post is accessible here.

Satya Nadella also backed the effort, saying open-weight models are essential to a healthy AI ecosystem and can strengthen U.S. competitiveness while expanding economic opportunity and protecting national security. His endorsement is available here.

The breadth of the coalition is notable given that the participating companies operate under different AI strategies. Meta has released its Llama family of large language models under open-weight licenses, making them freely downloadable for most use cases. OpenAI, by contrast, has built its business around proprietary models accessed primarily through its own API and ChatGPT interface. Microsoft partners closely with OpenAI while also hosting a range of models on its Azure cloud platform, and NVIDIA supplies the GPUs that underpin training and inference for both open and proprietary models across the industry.

The letter draws a parallel between today's AI debate and the rise of open-source software during the 1980s. It notes that open-source software became the foundation for much of the internet and now supports systems used by major technology companies, the U.S. military, and federal agencies. According to the document, open-weight AI models could play a similar role by enabling organizations to download, inspect, modify, and operate advanced models on their own infrastructure.

Letter Outlines Benefits and Risks

The document states that open-weight AI models allow startups, businesses, universities, and public institutions to leverage advanced AI without building frontier models from scratch or paying for every task through proprietary services. Organizations can select different models for different workloads while retaining control over their data, specialized capabilities, and accumulated knowledge. This is a distinguishing feature of open-weight models, where the trained parameters are published for download but the underlying training data, code, and methodology may not be fully disclosed, a narrower form of openness than the open-source software tradition the letter invokes.

The letter also addresses security concerns head-on. It acknowledges that released model weights are difficult to control or reverse after publication. However, it argues that restricting open-weight models is not the appropriate response, as cybersecurity defenders need access to advanced AI systems to identify threats, test vulnerabilities, and respond to attacks. The document's framing responds to a recurring concern among some AI safety researchers and policymakers that widely available powerful models could lower the barrier for malicious actors to generate harmful content or conduct cyberattacks.

Furthermore, the document states that transparency enables researchers and developers to examine model behavior, conduct benchmarking, perform red-team testing, identify vulnerabilities, and improve safeguards. It argues that concentrating advanced AI in a limited number of closed models could create single points of failure, underscoring the importance of a diverse and open AI ecosystem.