David Sacks, Co-Chair of the President’s Council of Advisors on Science and Technology, accused Anthropic of trying to stifle competition in the AI industry, warning that efforts to restrict open-source AI models could cripple the broader U.S. AI ecosystem.

Speaking on a recent episode of the All-In Podcast, Sacks argued that labeling Chinese open-source AI models as tainted by intellectual property theft would discourage American developers from building on publicly available models.

He said such a move would "basically put a dagger through the heart of the entire American open source ecosystem," adding that Anthropic "doesn’t want to have the competition."

Open Source AI Is ‘Not a Chinese Model Anymore,’ Says Sacks

Sacks said critics were overlooking how open-source AI development actually works. Once model weights are released publicly, he argued, developers around the world, including in the U.S., can legally download, modify and build on them using their own infrastructure, without sending data back to China.

To illustrate his point, Sacks pointed to examples including Mira Murati’s startup Thinking Machines and AI coding startup Cursor, saying both leveraged the open-source Chinese model Kimi K2.5 as a foundation before further training it on their own data.

Calling such models “tainted” because of unproven intellectual property allegations, he argued, would effectively prevent American companies from benefiting from open-source innovation and ultimately weaken the U.S. AI ecosystem.

Sacks Draws Line Between AI ‘Weights’ and ‘Outputs’

Sacks also argued that policymakers often conflate AI model weights, the underlying parameters that power a model, with the outputs those models generate. He said stealing proprietary model weights would amount to software theft, but maintained that using a model’s outputs to train another AI system, a process known as distillation, is fundamentally different.

He further contended that OpenAI and Anthropic have themselves defended training AI models on publicly available content, pointing to OpenAI’s legal defense in The New York Times copyright lawsuit.

According to Sacks, if AI companies argue they can learn from online content without copying it outright, then criticizing other companies for learning from AI-generated outputs presents an inconsistent standard.

He added that companies concerned about distillation should focus on strengthening their own safeguards and enforcing their terms of service rather than seeking broader restrictions on open-source AI.

Jensen Huang Joins the Open-Source AI Push

It also defends AI distillation as a standard development practice, urging policymakers to address legitimate intellectual property concerns through targeted legal frameworks rather than broad restrictions on open-source AI.

Disclaimer: This content was partially produced with the help of AI tools and was reviewed and published by Benzinga editors.

Photo courtesy: Shutterstock