
What happened
Writing after the 2026 midterms, an AI researcher says voters want an immediate, indefinite, international moratorium on AI development, not just limits on data centers. His 26-author paper proposes a 'hardwired' pause: stop making training chips and build 'inference-only' chips that can only run approved models.
Why it matters
A pause targeting chips rather than data centers would sidestep the objection that adversaries could keep developing AI in secret. The researcher argues the economic hit would be limited because existing models stay available and inference-only chips could run them faster and more efficiently.
Summaries like this, every morning.
The commentary situates its proposal in a political shift. It notes that across the U.S., candidates who were considered shoo-ins were caught off guard by a populist backlash to data center buildouts. In Texas, Greg Abbott, a popular governor in a state Republicans have held for 32 years, was unseated by Democrat Gina Hinojosa, who campaigned on an immediate moratorium on data center construction. Meanwhile, vulnerable incumbent Kelly Ayotte adopted the call for a moratorium and won her reelection bid. Of the 41 data-center related ballot measures, 40 went against the industry.
The author argues this backlash extends beyond data centers to AI development itself. He points to September's bill by Sen. Bernie Sanders and Rep. Greg Casar to ban superintelligence, which many dismissed as a messaging exercise, and to Rep. Yassamin Ansari's call for a pause on advanced AI development as top of the Democratic agenda. The author also cites polling that most Americans, including 70 percent of Democrats, support American AI companies pausing unilaterally until Congress acts.
The paper's central idea, a 'hardwired' pause, is presented as a response to the argument that adversaries would continue development in secret. By replacing existing AI chips with inference-only chips, the author says more than 99.9 percent of existing chips could be located and governed, making secret development extremely difficult due to the complexity of the technology and supply chains. He also argues the economic impacts would be limited by the continued availability of existing AI models, which inference-only chips could run faster and more efficiently. While AI companies' stock prices might decline, he suggests monetary policy could mitigate a resulting decline in spending, and a broader recession seems unlikely under current conditions.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
Victoria Song tested Siri Recap on the Apple Watch Ultra 4 for two weeks

In a Gallup and Boston Children's Digital Wellness Lab survey of 15,925 people aged 15–22 across 16 countries…

Databricks won Best AI Solution for Data Intelligence & Governance at the 2nd annual A-Team AI in Capital Mark…

A Zenn guide lays out a pre-migrate review: catch untracked files via git ls-files since git diff skips them…

VIDRAFT tested 20 widely used open-weight models with 8 simulated tools and 100 attack instructions; all 20 at…

The author runs four guards from settings.json — a PreToolUse hook that flags rm -rf, git push --force, git re…
