By Stephen Nellis
SAN FRANCISCO, July 24 (Reuters) - Nvidia, Microsoft and other tech heavyweights made a public case to lawmakers on Friday in favor of open-source AI models, wading into a debate roiling the business and policy worlds over who controls the powerful technology.
In a letter posted on X and also signed by two dozen companies and groups including Meta Platforms and IBM, Nvidia CEO Jensen Huang said that lawmakers should avoid "premature restrictions on open models that stifle competition or drive innovation overseas."
The letter adds to the growing debate about open-source models that are harder to regulate, such as Nvidia's own and those released in recent weeks by Chinese labs, and the closed-source models controlled by specific companies such as OpenAI and Anthropic.
In recent months, Silicon Valley business leaders have bristled at the cost of closed source models. The CEOs of Microsoft and defense contractor Palantir Technologies have publicly argued that open-source models their customers can run inside their own data centers will help control AI costs.
Tech leaders have also chafed at controls that OpenAI and Anthropic build into their models. Hugging Face, the AI coding collaboration site that was hacked by a rogue OpenAI model, this week said that it had to use a Chinese open-source model to defend against the attack because closed-source models have restrictions on use for cybersecurity work.
At the same time, U.S. lawmakers alarmed by the rogue OpenAI cyberattack proposed legislation that would require a "kill switch" for AI models, and President Donald Trump's administration is weighing sanctions on Chinese open-source model makers over alleged theft of U.S. closed-source technology.
The letter from Nvidia and other companies acknowledged the concerns about technology theft but argued they should be addressed "through targeted legal and commercial frameworks rather than sweeping restrictions."
"Relying solely on closed models is not inherently safe: they can be breached, misused, or fail in ways that outsiders cannot detect," the letter said. "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."
(Reporting by Stephen Nellis in San Francisco; Editing by Emelia Sithole-Matarise)