Former White House Crypto Czar David Sacks Warns US Could Lose AI Race
Highlights
- Kimi K3 ranked first on Frontend Code Arena, raising new questions about US AI leadership.
- David Sacks warned strict US rules could slow innovation while China advances frontier AI models.
- Kimi K3 offers 2.8 trillion parameters, million-token context, and planned open weights by July 27.
Former White House Crypto Czar David Sacks has warned that China’s Kimi K3 model has intensified pressure on America’s AI strategy after topping a key coding benchmark.
He said heavy regulation, data center limits, and federal pre-approval plans could slow US innovation while rival nations move faster in the race for advanced artificial intelligence systems worldwide.
Kimi K3 Tops Coding Benchmark
David Sacks said China’s Kimi K3 has taken the top position on the Frontend Code Arena. He described the result as concerning because the model also scored near frontier levels on other tests.
Kimi K3 is built with 2.8 trillion parameters and a one-million-token context window. The model supports native multimodal use and is designed for long coding tasks and agent workflows.
Moonshot AI said Kimi Delta Attention can deliver up to 6.3 times faster decoding in million-token contexts. The company also said Attention Residuals improve training efficiency by about 25% at less than 2% extra cost.
Kimi K3 is already live on Kimi.com, Kimi Work, Kimi Code, and the Kimi API with open weights are expected by July 27.
David Sacks Criticizes US AI Rules
The former White House crypto czar, Sacks argued that the United States risks slowing itself through strict AI rules. He criticized efforts to restrict data center construction, expand state-level rules, and create federal review systems for frontier models.
The debate comes as the U.S. approved limited access to Anthropic’s Claude Mythos 5 for about 100 businesses and federal agencies.
In a letter to Anthropic co-founder Tom Brown, Commerce Secretary Howard Lutnick said,
“I have determined that appropriate safeguards are in place to permit certain trusted partners to access the Claude Mythos 5 Model.”
Sacks wrote, “This is how you lose the AI race.” He argued that other countries would not follow U.S. limits if America slowed its own AI development through heavier regulation.
He said the United States won the internet era through permissionless innovation. Sacks added that America could compete in AI the same way while addressing safety risks with targeted rules.
Benchmark Results Fuel Competition Debate
This development comes as President Donald Trump continues to try to make the US the “Crypto Capital of the World.” On the regulation side, the US has lagged, with Donald Trump urging the Senate to pass the CLARITY Act to avoid China overtaking the US in the crypto sector.
Meanwhile, China’s Kimi K3 also reached an Elo rating of 1668 on GDPval v2. That score was above GLM-5.2, GPT-5.5, and Claude Opus 4.8, though it remained below Claude Fable 5.
The model also reached 53% on AutomationBench-AA, taking the top position on that agentic SaaS workflow test. Concurrently, according to the report, on nextjs.org/evals, Kimi K3 has ranked ahead of Fable and posted comparable success in less time.
Sacks said AI safety should still be addressed, but he warned against rules that block infrastructure and model development. His remarks added to the debate over AI regulation, data centers, and U.S.-China competition.
If you want more control over your crypto portfolio, utilizing real AI crypto trading bots can help track market moves and support faster decisions.











