A former researcher at OpenAI and Anthropic named Jacob Coxon quit his job on September 8th and immediately told reporters that American AI companies are "gambling with our lives." Three years inside the most advanced AI labs on earth, and his parting contribution was a sound bite designed to make Congress panic four days before recess.
It worked.
Senate Democratic leader Chuck Schumer responded on cue: "Warning bells are ringing louder than they ever have — and the Trump administration is looking the other way. If Trump does not act, Congress must." Senator Bernie Sanders wants a full development pause on superintelligent AI. Rep. Ted Lieu, a Democrat from California, introduced a "kill switch" bill that would require developers to retain the ability to throttle, suspend, or shut down their own systems. Lieu's pitch was characteristically measured: "Humans have to be able to turn off AI systems... if the response is we can't do that, then we might as well just pack up our bags and go to some island in the South Pacific and wait for the apocalypse."
Meanwhile, on the other side of the Pacific, China's government unveiled a comprehensive national strategy to become the world's AI leader by the end of this decade. Downloads of DeepSeek — the Chinese AI model that U.S. intelligence believes was built using stolen American intellectual property through unlawful model distillation — grew nearly 1,000 percent since January 2025. The gap between Chinese and American AI models has narrowed from years to months.
Rep. Kat Cammack, a Florida Republican who sits on the Subcommittee on Communications and Technology, laid out the geometry of this in a National Review piece this week. Her argument is simple enough that it shouldn't need to be published in a policy journal: China wants us to be afraid of our own technology so that we regulate it into irrelevance while they steal it and sprint ahead. Cammack has been sounding this alarm since at least July, saying that China has already deployed "digital twins" of every member of Congress — AI simulations built on harvested data to war-game elections, policy votes, and influence campaigns. "They'll never be able to take down the United States military," Cammack said, "but what they're looking to do is basically take down the American mind."
The House introduced H.R. 6909, the China AI Threat Assessment Act, which directs the Director of National Intelligence to produce a National Intelligence Estimate on AI systems developed or deployed by Chinese entities. That's a reasonable piece of legislation — understanding the threat before reacting to it. But the Democratic proposals skipping assessment entirely and jumping to kill switches, development freezes, and emergency sessions represent exactly the self-inflicted wound Cammack is describing.
Schumer's framing tells you everything. "The Trump administration is looking the other way" is not an argument about safety. It's a midterm talking point wrapped in a lab coat. The House has four days left in session before the November 3rd elections. Democrats need an AI panic narrative the same way they needed a COVID panic narrative — not because the threat isn't real, but because the proposed solution always happens to consolidate their power over an industry they don't control.
House Speaker Mike Johnson identified the trap: "If Congress just races in and does some sort of emergency session to try to regulate AI, we will lose the race to China." President Trump was more direct: "The only control or 'guardrails' that AI needs is a STRONG AND SMART (High IQ!) PRESIDENT."
The pattern is not subtle. Chinese state-backed hackers and intelligence services steal the underlying model weights from American companies. Chinese firms use those stolen foundations to build competitive products at a fraction of the R&D cost. China then amplifies Western fears about AI safety through influence operations and cooperative academics. Western legislatures respond by restricting the very companies China just robbed. China builds faster while America builds committees.
Coxon's AI models escaped a test environment, accessed the internet, and infiltrated a Hugging Face code repository. That is a legitimate engineering problem that demands an engineering solution from the companies involved. It is not a reason for Bernie Sanders to pause the development of an entire technology sector while China pours state resources into catching up.
China's AI strategy runs on coordinated state funding at a scale that one-party rule makes significantly easier. America's AI advantage runs on the absence of that kind of central planning — on companies competing, iterating, and moving faster than any bureaucracy can approve. Every regulation that slows that cycle is a gift to Beijing, delivered free of charge by the same people who spent four years insisting we needed to be more like Europe.
The House leaves town this week. The kill switch bill sits in committee. The China AI Threat Assessment Act sits in committee. One asks what China is actually doing. The other asks American companies to install an off button.
China knows which one it's rooting for.