SAN FRANCISCO, California: Nvidia, Microsoft and dozens of technology companies and industry groups have urged U.S. lawmakers to support open-source artificial intelligence models, arguing that broad restrictions could hurt competition and push innovation outside the United States.

The appeal came in a letter posted on X by Nvidia CEO Jensen Huang and signed by more than two dozen companies and organizations, including Meta Platforms and IBM, as policymakers continue to debate how advanced AI systems should be regulated.

The discussion has intensified as companies and governments weigh the risks and benefits of open-source AI models, including those developed by Nvidia and recently released by Chinese laboratories, against closed-source models controlled by companies such as OpenAI and Anthropic.

In recent months, business leaders in Silicon Valley have raised concerns about the cost of closed-source AI models. The chief executives of Microsoft and defense contractor Palantir Technologies have also argued publicly that open-source models, which customers can operate within their own data centers, could help reduce AI costs.

Technology companies have also criticized restrictions built into closed-source models. Earlier this week, Hugging Face, the AI coding collaboration platform that was hacked by a rogue OpenAI model, said it relied on a Chinese open-source model to defend against the attack because closed-source models imposed restrictions on cybersecurity-related work.

The debate comes as U.S. lawmakers, responding to the rogue OpenAI cyberattack, have proposed legislation requiring a "kill switch" for AI models, while President Donald Trump's administration is considering sanctions against Chinese developers of open-source AI models over allegations they stole U.S. closed-source technology.

In their letter, the companies acknowledged concerns about technology theft but said the issue should be handled through legal and commercial measures rather than broad limits on open-source AI.

The letter also said: "Relying solely on closed models is not inherently safe: they can be breached, misused, or fail in ways that outsiders cannot detect. Open weight models, on the other hand, allow a broad community of researchers and developers to examine their behavior, identify vulnerabilities, develop safeguards, and improve them over time."