A world where the most available models are the hardest to trust
4 July 2025 (Interconnects AI). Nathan Lambert, in a 2025 Interconnects essay proposing 'The American DeepSeek Project,' warned about a future dominated by powerful closed American models alongside widespread open Chinese models.
More from Nathan Lambert · AI and politics · AI and security · 2025
Who else?
What it refers to: A future in which extremely powerful closed American models counterbalance a sea of strong, ubiquitous open Chinese models, which the essay ties to the Chinese Government.
Other voices
- “Present multiple threat vectors” US DHS notice, Federal Register,
- “Enslaved to PRC-mediated AI” JD Vance, Politician,
- “Openness could become a structural advantage for China's AI industry” Zhang Yaqin,
- “We are at the end of open source models now” Tim Dettmers,
- “Power isn't concentrated in the hands of a small number of companies” Mark Zuckerberg, Builder,
- “China should gradually become a contributor instead of freeriding” Liang Wenfeng,










