Several major AI companies in the U.S. have co-signed an open letter advocating for unrestricted access to open-weight AI models. This letter emphasizes the importance of openness in AI to drive innovation and accessibility.
Nvidia CEO Jensen Huang initiated the campaign by tweeting a statement that champions open-weight AI models, which can be downloaded, viewed, and edited by anyone. The letter argues that such models are essential for making AI “more accessible, adaptable, and widely available.” Companies including Meta, Microsoft, Perplexity, and IBM have also backed this initiative, with OpenAI joining later despite its previous focus on closed models.
The letter comes amid heightened scrutiny from the U.S. government regarding AI regulation. Recent efforts led by President Trump to evaluate AI models and determine their release timing have been viewed as potential threats to innovation. The signatories argue against what they consider “premature restrictions” by governments.
Notably, Anthropic and Google did not co-sign the letter. Their absence raises questions, particularly in light of accusations against Chinese company Moonshot AI for intellectual property theft. Moonshot AI is alleged to have distilled Claude’s closed Fable 5 model to develop its open Kimi K3 model, which is set for release on Monday.
Moonshot AI has denied these allegations, claiming that Kimi K3 will have significant potential as an open-weight model. The U.S. government, however, has responded by labeling this activity as IP theft, with officials issuing strong warnings against such practices.
U.S. Treasury Secretary Scott Bessent stated, “We support open-source AI and the innovation it unlocks. But open source is not open season on American IP. When PRC firms conduct covert, industrial-scale distillation attacks that cross the line into IP theft, sanctions and Entity List designations will be on the table.”
