OpenAI recently signed an open letter supporting the development of open-weight AI models, signaling a notable change in its approach. This move aligns the company with a Microsoft-led initiative advocating for open AI models to bolster US leadership in the field.
The letter, titled “Open Weights and American AI Leadership,” has garnered support from 25 organizations such as NVIDIA, Meta, Palantir, and Hugging Face. It stresses that open-weight models are key to maintaining competitive advantage, improving safety, and speeding up innovation within the United States.
Previously, OpenAI had focused on closed systems, but in August 2025 it launched the gpt-oss family of models, including gpt-oss-120B and gpt-oss-20B, licensed under Apache 2.0 for local use on consumer devices. These models are accessible on platforms like Hugging Face and GitHub, reflecting a gradual pivot toward openness.
The letter also highlights concerns about national security and competition, particularly as Chinese AI labs release advanced open-weight models. The signatories argue that limiting open-weight development domestically would hand strategic advantages to foreign competitors.
Behind the public letter lies an active policy debate in the US over restrictions on open-weight models, especially those originating from China. The letter points out that open models enable independent audits, helping to uncover vulnerabilities and enhance cybersecurity.
NVIDIA CEO Jensen Huang amplified the campaign by posting about it on X, marking his first engagement on the platform. Interestingly, major AI players like OpenAI, Anthropic, and Google did not sign the letter.
For the crypto and decentralized AI communities, this development is significant. Open-weight models are foundational for decentralized AI, as they expand the market for crypto-native AI infrastructure. OpenAI's gpt-oss models, which can run on consumer hardware, support the idea that distributed networks of consumer GPUs might compete with centralized cloud services.
