Why Nvidia, Microsoft and Meta want the US to support open-weight AI | Technology News

A group of leading technology companies, including Nvidia, Microsoft, Meta, IBM and Andreessen Horowitz, has urged the US adminration to support the development of open-weight artificial intelligence (AI) models. In a joint letter signed 25 firms, the coalition argued that open-weight AI is key to maintaining America’s technological leadership and warned against premature restrictions that could limit innovation. The appeal comes as Washington mulls tighter controls on advanced AI technologies amid growing concerns over the rapid progress of Chinese AI companies.Nvidia CEO Jensen Huang, a vocal proponent of open-weight AI, shared a post along with the signed letter. “I’m sharing a letter Nvidia signed on why open models matter. AI will transform every industry, power every company, and be built every country,” he wrote.
Huang said open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty. “The world needs both frontier closed models and frontier open models,” he added.
The letter comes amid a brewing debate in Silicon Valley over the rapid spread of Chinese AI tools, especially open-weight AI models that have shown they can compete with or surpass some of the best closed frontier models developed in the US.
This is the second letter this week urging the Donald Trump adminration to adopt a more favourable approach towards open-weight AI models. On July 22, a group of over 200 startups under the banner of the Little Tech Association wrote to Michael Kratsios, science adviser to President Trump, and US Commerce Secretary Howard Lutnick over the outright ban. The group argued that restricting access to AI models abroad could weaken US startups and create a monopoly among AI giants.
Why companies are backing open-weight AI
With AI becoming the next frontier of technological competition, two dinct approaches seem to be emerging. While companies such as OpenAI, Anthropic, and Google continue to build proprietary AI systems, another group is pushing for wider adoption of open-weight models. The letter urging the US government to support open-weight AI models as part of the national strategy has been signed Meta, Microsoft, IBM, Nvidia, Hugging Face, Dell, Mozilla, Box, ServiceNow, CrowdStrike and others.
Why are some of the biggest technology companies in the world calling on policymakers to accept AI models that anyone can download, inspect, and customise? To understand, it is important to define open-weight AI.Story continues below this ad

Unlike closed AI systems, open-weight models allow developers and businesses to download the model’s trained parameters—or ‘weights’—and run them on their own infrastructure. This gives organisations greater control over how the model is used, modified, and deployed without depending entirely on the company that created it.
In essence, this idea is similar to the open-source software movement that transformed computing decades ago. Just as Linux became the foundation for much of today’s internet, supporters of open weight believe AI could become the building block for the next generation of AI applications. Unlike fully open-source AI, open-weight models typically make the trained model weights available without necessarily releasing the underlying training data or source code.
Make AI cheaper and more accessible
Training advanced AI models can cost billions of dollars, putting frontier AI out of reach for most startups, governments, and universities. The letter argues that open-weight models can address this challenge allowing developers to build on exing AI models instead of creating one from scratch. Companies can also choose a model best suited to their needs, there reducing costs while reserving the most powerful—and expensive—models for tasks that truly require them. For smaller firms, this lowers the barrier to entering the AI race.
Perhaps the biggest argument in favour of open-weight AI is competition. If the world’s most capable AI models are owned a small group of companies, they could likely control pricing, access, and innovation. As per the letter, open-weight models encourage competition not only among AI developers but also across cloud providers, chipmakers, software companies and AI application developers.Story continues below this ad
Greater competition, the authors say, leads to faster innovation, lower costs, and wider economic benefits. This is particularly important as AI becomes part of everyday business software, healthcare, manufacturing, education and scientific research.

Another key advantage of open-weight AI is that it gives greater control to businesses.  This is crucial as most businesses investing heavily in AI do not want to be dependent on a single vendor. With open-weight models, companies can keep their own data, customise models for specialised tasks and deploy them wherever they choose. They can also continue improving the model over time without relying entirely on one AI provider. According to the letter, this helps organisations retain the value they create through their AI investments.
Behind the debate lies a commercial reality. Open-weight AI lowers the cost of building AI products, allowing startups and smaller companies to develop applications without spending billions on training frontier models. At the same time, companies offering proprietary AI platforms—such as OpenAI, Anthropic and Google—have a financial incentive to keep their most advanced models exclusive. This has fuelled a broader debate over whether the future of AI should be dominated closed systems or supported a mix of proprietary and open-weight models.
What about security risks?
One of the biggest concerns around open-weight AI is a technique known as dillation, in which a smaller model learns from the outputs of a larger and more capable one. Last month, Anthropic accused Chinese tech giant Alibaba of stealing its IP through dillation attacks. This week, the White House alleged that Chinese startup Moonshot AI developed its Kimi K3 model dilling Anthropic’s Claude Fable 5 model.Story continues below this ad
Most of the criticism of open-weight AI centres on the risk if it falls into the hands of bad actors. Critics believe that allowing advanced AI models to be widely accessible would make them prone to misuse. The tech companies that signed the letter acknowledged the risks but argued that openness can also strengthen security.
A recent example was a cybersecurity incident involving Hugging Face. After an autonomous AI agent linked to OpenAI breached the platform during a security evaluation, Hugging Face said it initially relied on hosted frontier models for forensic analysis but was blocked their safety guardrails. The company instead used the open-weight Chinese AI model GLM-5.2 on its own infrastructure to investigate the breach, arguing that open-weight models may also strengthen cyber defence allowing security teams to analyse sensitive data locally.

The letter says cybersecurity defenders need access to advanced AI to counter increasingly sophicated attacks. Open-weight models, it argues, allow more researchers to identify weaknesses, improve safeguards and independently test systems rather than relying solely on the original developer. The letter compares this to open-source software, where public scrutiny has often helped uncover and fix vulnerabilities faster.
Instead of restricting open-weight AI, these companies want the US government to expand access to computing resources, invest in shared datasets and evaluation tools, and avoid regulations that may make it difficult to develop or use open models. They also warn against broad restrictions on AI training techniques such as dillation, arguing that legitimate improvement methods should not be confused with intellectual property theft.Story continues below this ad
The debate over open versus closed AI is becoming one of the defining questions of the AI era. While closed models promise tighter control, open-weight models aim to spread AI capabilities across business domains, researchers, and developers. The coalition argues the latter approach will help the US maintain technological leadership encouraging innovation, increasing competition, and ensuring AI benefits more sectors of the economy rather than a few companies. The outcome could eventually shape not only the future of AI in the US but also who leads the next phase of global AI development.