AI researcher Jacob Coxon’s epically viral post on X this week — warning that the leading labs were not addressing the foreseeable, growing risk of catastrophe — reignited the AI safety debate in Washington, launching 1,000 op-eds and bipartisan demands for action. While Coxon was hardly the first to warn of AI doom, nor the most prominent, his post struck a chord, with many proclaiming a “hinge moment”. Within days Sam Altman ruled out a 2026 IPO and Dario Amodei published “We Must Pace the Frontier,” with Altman and Musk endorsing the call for slowing research.
From nuclear weapons to nanotechnology, technology catastrophists have long warned of immense risk from emerging technologies. While many such warnings proved overwrought or wildly-premature, others catalyzed support for constructive safeguards that averted bad outcomes.
AI regulations are likely a matter of when not if. But what to regulate, how, overseen by whom and to what effect remain unclear… likely driven by a crisis, US-China cooperation, industry-led frameworks and/or the 2028 Presidential campaign.
1. Coxon is neither the first nor the most prominent technologist to warn of AI risk.
2. Experts have often warned of risk from new technologies.
3. While many urge a pause in frontier research due to demonstrated risks of misaligned machines escaping control, others see peril in failing to maintain U.S. global AI leadership, especially vs China.
4. Urgent warnings can help galvanize constructive safeguards… but action often only comes when danger becomes tangible, measurable or impossible to ignore.
5. Neither AI nor data centers crossed the 4% mark (of voters naming their top issue) in July… but that was before the Hugging Face hack, Coxon post & data center ad blitz… will exit polls tell a different story?
6. What’s Next?
Recommended additional reading:
Jim VandeHei & Mike Allen, “It’s Not Too Late,” Axios Behind the Curtain, Sept. 12, 2026
Peggy Noonan, “Pause AI for Humanity’s Sake,” Sept. 10, 2026
Derek Thompson, “AI Safety Is Having a Moment,” Sept. 10, 2026
All In Podcast, “AI Kills Everybody or Doomer Psi-Op?,” Sept. 11, 2026
PODCAST
Always great joining Chuck Todd to talk data centers, campaigns & the economy.
The reality is that innovation cannot stop, especially one-sided innovation that constrains Americans and not the Chinese. But trying to stop AI's advancement would be like Roosevelt trying to stop nuclear research in the hope that our enemies wouldn't have suddenly developed a weapon themselves. The key will be how we regulate a capable AI that could, if not contained, be highly destructive. Like the US and the USSR, the most dangerous powers will have to come together to voluntarily restrain their own destructive capabilities, knowing that mutually assured destruction applies as much to AI as it did to nuclear weapons. It won't be altruistic. It will be out of necessity and the will to live. Then we have to stop and think about whether there are certain countries we don't want to have the capacity for the most capable AI systems. AI non-proliferation might be a major issue in the future. And it might be a lot harder than nuclear non-proliferation.
If it can be built, it will be built. You can’t ‘pause’ or stop innovation. Somebody will build it, and it will be a tool that can be used for good or evil. We pray good wins.