Jensen Huang Joins Microsoft and Meta in Urging Congress Not to Stifle Open-Source AI

NVIDIA (NVDA-US), Microsoft (MSFT-US), Meta Platforms (META-US), IBM (IBM-US), and other major tech firms jointly sent a letter to the U.S. Congress on Friday, publicly supporting open-source AI models. They urged lawmakers to avoid imposing early or overly broad restrictions on open-source models, warning that such measures could stifle market competition and drive AI innovation overseas.

The open letter was spearheaded by NVIDIA CEO Jensen Huang and signed by approximately 24 companies and industry organizations. In the letter, Huang stated that governments should avoid "imposing premature restrictions on open-source models, thereby stifling competition or pushing innovation overseas."

This joint effort highlights the intensifying debate within the AI industry over the direction of open-source versus closed-source models. Currently, models released by NVIDIA and several Chinese AI labs adopt relatively open-weight or open-source strategies. In contrast, companies like OpenAI and Anthropic primarily rely on closed-source models, where the enterprise controls model access and updates.

In recent months, Silicon Valley companies have grown increasingly dissatisfied with the rising costs of using closed models. Senior executives from Microsoft and Palantir Technologies (PLTR-US) have publicly stated that allowing enterprises to deploy open-source models in their own data centers can reduce AI usage costs, decrease dependence on a single AI supplier, and enhance corporate autonomy.

Beyond cost considerations, multiple tech firms have questioned the usage restrictions embedded in models developed by OpenAI and Anthropic, arguing that certain safety mechanisms actually hinder legitimate corporate applications of AI technology.

AI development platform Hugging Face recently reported being targeted by a cyberattack launched by a runaway AI system derived from an OpenAI model. Due to usage restrictions on closed models for security purposes, the company ultimately used a Chinese open-source AI model to assist in defense, highlighting the flexibility advantage of open-source models in cybersecurity research.

On the other hand, the U.S. Congress has heightened its alert over AI-driven cyberattacks. Some lawmakers have proposed draft legislation requiring AI models to include an "emergency shutdown mechanism (kill switch)" to allow the government to terminate model operations when necessary. Meanwhile, the Trump administration is evaluating whether to impose sanctions on certain Chinese open-source AI model developers, citing alleged theft of U.S. closed-model technologies.

In response, NVIDIA and other signatories acknowledged that the U.S. must take seriously the risks of AI technology theft and national security threats. However, they argued that these issues should be addressed through more precise legal and commercial mechanisms, rather than blanket restrictions on open-source models.

The letter emphasized that reliance on closed models does not inherently guarantee greater security, as closed models can also be hacked, misused, or suffer undetectable errors. In contrast, open-source models allow a broader community of researchers and developers to inspect model operations, identify vulnerabilities, build protective measures, and continuously improve overall security through iterative enhancements.

In recent years, with Meta launching its Llama series models and Chinese firms like DeepSeek releasing high-performance open-source AI models, the global AI industry has gradually split into two major camps: open-source and closed-source. Market analysts believe that this competition over technology, security, regulation, and business models will not only shape corporate AI strategies but also determine the future trajectory of the global AI industry.