Industry leaders warn against early curbs on open-weight AI

Jensen Huang posted a letter online, signed by 25 tech giants including Meta, Microsoft, Perplexity, IBM and Dell, urging lawmakers to advocate for open-weight AI models over closed-weight alternatives.

Huang cautioned against “premature restrictions on open models,” stating that it stifles competition and drives AI innovation to other countries.

The number of people signing the letter increased to 50 after the Nvidia CEO published it on X.

Policymakers remain divided over how to regulate open-weight AI models, including those from China, compared with models developed by companies such as OpenAI and Anthropic.

According to the coalition, closed models can still be breached, whereas open models let researchers around the world find weaknesses and build better safeguards.

Open-weight AI models allow users to download, modify, and deploy them locally. By comparison, leading US large language models, including Claude Fable 5 and the GPT-5.6 series, remain closed-weight and proprietary.

Popular Chinese AI systems, including DeepSeek and Moonshot AI’s Kimi K3, are built as open-weight models.

The appeal comes as rising costs and security concerns reshape the AI landscape. Microsoft and Palantir executives say open models enable customers to run AI in their own data centers, cutting costs.

Meanwhile, cybersecurity risks have also come into focus. Earlier this month, Hugging Face reported that it used an open-weight Chinese model to defend against a rogue OpenAI model after Anthropic’s Fable 5 failed to determine whether Hugging Face was the attacker or the target.

After the incident, a bipartisan AI Kill Switch Act was introduced in the House.

The Trump administration is also looking at sanctions targeting Chinese open-weight AI developers over allegedly stealing American IP.