In March we saw AI become a mainstream political issue and without a superior strategy Republicans will lose power.
Sens. Bernie Sanders and Alexandria Ocasio-Cortez have taken the Pause AI position. Geopolitics clamped energy flow to Taiwan, and helium used to finish chips.
Sooner or later a deal must be made with China to determine the fate of humanity, it can be defined by a Gavin Newsom/Cortez presidency or a Trump/Vance presidency. Donald Trump says “there should be” an AI kill switch. The world is in a chaotic state, conditions are perfect.
Trump’s energy strategy creates opportunity, Taiwan needs U.S. natural gas and helium. The 2025 AI policy produced massive investment of 2% GDP and enough hardware to cook an intelligence explosion, even if Taiwan were lost. Sen. Sanders’ proposed moratorium on new AI data centers is late.
The post-war energy reality means the U.S. Government takes a bigger role in buying AI services, companies will divert more chips from consumers to training, with new AIs kept private or classified, Anthropic’s Mythos model proves this.
Policy around AI becomes heavier than the Oppenheimer debates after World War II.
There’s a meme where Cillian Murphy as Oppenheimer looks shattered, responsible for permanently increasing global risk in the name of peace. It’s relatable for an AI safety researcher or policy advocate. The Machine Intelligence Research Institute founded by Eliezer Yudkowsky wishes to refuse like Einstein, with a proposed treaty with China that would ban all training runs beyond what Anthropic used 2 years ago. Senator Sanders is indirectly championing that proposal.
We also have Connor Leahy’s Control AI organization, he cut his hair and moved to D.C. after spending 4 years working on, and failing to, solve the problem of interpreting these things. Looking inside the machine is a hard enough technical problem, he thought politics would be easier!
Pausing AI is politically complex and technically simple (on paper), controlling AI is politically sellable (governments like control) and technically complex. The UK’s AI Safety Institute published a framework called “Control Arena” as a tool to create new control protocols to test in a variety of Arena simulations.
We could have millions of bright people working in AI Control, absorbing some of the lost software engineering jobs. AI Control is a sober middle-ground.
On the right we have White House AI Action Plan co-author Dean Ball who questions how authoritarian a world with a Pause AI treaty would have to be. Having dealt with bad banking policies as a Bitcoiner for many years, I can attest that an OECD-wide regime of global control like Common Reporting Standards can create this “New World Order” authority and be semi-effective at limiting research.
Dean has become more concerned about risk since last year, echoing Trump’s comment to Maria Bartiromo that “there should be” an AI kill switch.
The center-left policy wonk Noah Smith recently went on Doom Debates, hosted by Pause AI advocate Liron Shapira, and criticized their failure to sell D.C. on policy. Smith correctly pointed out that policy has *never* been made based on the unprecedented.
Nuclear regulation for instance was prompted by events with limited damage. That’s why I describe near-term risks based on limited real damages and advocate for defensive policies that also help with sci-fi risks.
The “6th generation warfare” I’ve described is incremental to 5th generation warfare that intelligence analysts and White House strategists understand, supplementing the strategic planning of think tanks like Paule Scharre’s CNAS or the new State Department Bureau of Emerging Threats.
With this political landscape in mind, I have 90%+ confidence that an alive humanity in 2030 will have a U.S.-China AI grand bargain in two possible forms: Democrat or Republican.
The Democrat form comes from Newsom negotiating politely and looks like Biden-era crypto policy applied to AI: caps, soft and hard criminalization, talent flight.
Maybe it would keep us safe, but it would mean worldwide Chinese-style surveillance.
The Republican form is negotiated with the leverage of an ongoing trade/cold war and acknowledges: 1) keeping the best AI classified and below superintelligence is best for national security, 2) kinetic war to enforce the treaty is the new mutually assured destruction, 3) decentralized AI is a 1st/2nd Amendment right and a lid on frontier AI means capabilities catch up 80%, but limited by slowness and chips, to really compete decentralized AI needs massive networking.
Trump’s entire legacy is tied up into whether or not his AI policy leads to a China treaty, and 2026 is actually a great year for it.
You don’t want Bernie influencing mid-terms based on a narrative of jobs and survival. You want the issue resolved with a binding grand bargain that leaves room for the U.S. Constitution.
Patrick Dugan is an independent AI researcher and founder of MoralityLab.
The views and opinions expressed in this commentary are those of the author and do not reflect the official position of the Daily Caller News Foundation.
All content created by the Daily Caller News Foundation, an independent and nonpartisan newswire service, is available without charge to any legitimate news publisher that can provide a large audience. All republished articles must include our logo, our reporter’s byline and their DCNF affiliation. For any questions about our guidelines or partnering with us, please contact [email protected].
