US think tank proposes military strikes to block China's AGI
A Washington think tank just put missiles on the table in the AI race — and Beijing's state media answered with threats of its own. The lead story this morning is the darkening of what was, until this week, a benchmark-and-chips rivalry.
A former Obama administration national security official has called on the US to prepare for military strikes on Chinese data centers to stop Beijing reaching AGI first. Jacob Stokes, deputy director of the Indo-Pacific Security Program at the Center for a New American Security, lays out the escalation ladder in a report titled "Superpowers and AGI." He starts with espionage and offers a "justified" case for stealing Chinese AGI tech, moves to offensive cyber operations, and lands on the "final option" — kinetic attacks, in his words "destroying things with missiles and bombs." He told an online event that the Department of Defense and the National Security Agency should begin assessing what intelligence they would need to justify such measures, working backwards from the technology to the policy "in the same way that in a past era, policymakers would learn about nuclear weapons."
The framing is deliberately nuclear-adjacent, and the reaction was immediate. Hu Xijin, the former editor-in-chief of China's state-backed Global Times, responded on social media that if the US dared strike Chinese data centers, retaliation "would directly target the US mainland with missile strikes." That exchange lands a day after OpenAI declared "Welcome to the AGI era" at the GPT-6 Astra launch, and weeks before Trump and Xi are set to meet in Washington on September 24, with AI governance confirmed on the agenda. Arms analysts aren't buying Stokes's calculus — William Hartung of the Project on Government Oversight called the idea "the height of recklessness," warning it risks "a shooting war between two nuclear-armed powers."
The honest read: this is more provocation than policy. Stokes assumes AGI is inevitable even though no such system exists, he treats America's roughly seven-month frontier lead as a position to defend rather than a sign the premise is shaky, and he imagines a nuclear-arms-race framing for a technology that behaves nothing like a weapon that can be counted. The more important signal is that mainstream security figures now feel free to say the quiet part about compute — that AI infrastructure itself has become a military target. That is a genuinely new threshold. Watch whether the Trump-Xi summit produces anything that walks it back.
Gansu province is putting AI on every school timetable in China's latest state-driven push to build AI literacy from the ground up. In an action plan covering 2026 to 2030, the provincial education department and four other agencies will require elementary school upper grades, junior high, and high school to schedule at least eight class hours of AI each school year starting autumn 2027, with general AI education reaching every school in the province. Higher education gets around ten provincial AI general courses open to all majors, and vocational programs run AI coursework across the whole middle-to-upper vocational track. Students' AI literacy gets folded into a digital evaluation file, and the plan targets building more than 45 AI education base schools and over 1,000 typical application scenarios by 2028. It's the kind of top-down mandate that has become China's signature approach to AI — and a sharp contrast with the US, where AI education policy is still being fought over school by school and state by state.
What to watch: Whether the Trump-Xi summit on September 24 turns any of this AGI-race talk into actual governance, and whether other Chinese provinces follow Gansu's curriculum mandate.
Should a US national security establishment be planning for kinetic strikes against compute centers — and would that make any nation less likely to race toward AGI? Tell us in the comments.
