David Sacks, Co-Chair of the President’s Council of Advisors on Science and Technology, has accused Anthropic of attempting to stifle competition by restricting open-source AI models. During a recent episode of the All-In Podcast, Sacks warned that characterizing Chinese open-source models as tainted by intellectual property theft would harm the broader U.S. AI landscape. He stated that labeling these models as compromised would effectively put a dagger through the heart of the American open-source ecosystem, arguing that Anthropic seeks to eliminate competitive pressure. Sacks clarified that once model weights (underlying parameters powering a model) are released, developers globally can modify them locally without sending data back to China. He cited startups like Thinking Machines, led by Mira Murati, and Cursor as examples of companies that built upon the Chinese Kimi K2.5 open-source model to develop their own systems.

Sacks further criticized the industry's stance on distillation, where models are trained using the outputs of other AI systems. He contended that if companies like OpenAI defend their own training on publicly available content, then criticizing others for using AI-generated outputs creates an inconsistent standard. Instead of pursuing broad restrictions on open-source AI, Sacks suggested companies should focus on enforcing their existing terms of service. His remarks align with a growing movement, evidenced by a letter signed by over 20 organizations, including Nvidia, Meta, Microsoft, IBM, Palantir, Hugging Face, Mistral AI, and Perplexity. Nvidia CEO Jensen Huang recently used his first-ever post on X to support this coalition, asserting that open-weight AI is vital for U.S. technological leadership. The group argues that restrictions on open models would hinder innovation, reduce cybersecurity capabilities, and limit the accessibility of AI for researchers and startups.

David Sacks, Co-Chair of the President’s Council of Advisors on Science and Technology, has accused Anthropic of attempting to stifle competition by restricting open-source AI models. During a recent episode of the All-In Podcast, Sacks warned that characterizing Chinese open-source models as tainted by intellectual property theft would harm the broader U.S. AI landscape. He stated that labeling these models as compromised would effectively put a dagger through the heart of the American open-source ecosystem, arguing that Anthropic seeks to eliminate competitive pressure. Sacks clarified that once model weights (underlying parameters powering a model) are released, developers globally can modify them locally without sending data back to China. He cited startups like Thinking Machines, led by Mira Murati, and Cursor as examples of companies that built upon the Chinese Kimi K2.5 open-source model to develop their own systems.

Sacks further criticized the industry's stance on distillation, where models are trained using the outputs of other AI systems. He contended that if companies like OpenAI defend their own training on publicly available content, then criticizing others for using AI-generated outputs creates an inconsistent standard. Instead of pursuing broad restrictions on open-source AI, Sacks suggested companies should focus on enforcing their existing terms of service. His remarks align with a growing movement, evidenced by a letter signed by over 20 organizations, including Nvidia, Meta, Microsoft, IBM, Palantir, Hugging Face, Mistral AI, and Perplexity. Nvidia CEO Jensen Huang recently used his first-ever post on X to support this coalition, asserting that open-weight AI is vital for U.S. technological leadership. The group argues that restrictions on open models would hinder innovation, reduce cybersecurity capabilities, and limit the accessibility of AI for researchers and startups.