Q&A with Matt Sheehan, a senior fellow at the Carnegie Endowment for International Peace, on US-China AI race, distillation, US chip export controls, and more
THE SO WHAT
The US–China AI race is increasingly about distillation and export controls, not just frontier training runs—who can do more with constrained hardware. Multinationals need a dual-stack strategy: one architecture that assumes full access to advanced chips, and another that can operate under tighter export and supply limits.
READ THE SOURCE
MORE FROM THE WIRE
Applied AICloudflare: Stay discoverable in search while disallowing AI training
Search and AI crawling are being decoupled at the infra layer—Cloudflare is giving sites a way to stay in Google while opting out of model training. If you run a content-heavy property, expect your SEO, licensing, and AI strategy to become one negotiation instead of three.
Mark Zuckerberg says AI labs can slow down on their own when safety demands it
Treat "we can slow down ourselves" as a signal that major labs expect to negotiate with regulators from a position of self-claimed responsibility, not constraint. If you're deploying on top of their stacks, assume safety posture will be driven by internal evaluators and PR pressure more than hard law for the next few years.
Applied AIWe don’t need AI regulation — leave safety to us, Nvidia’s Jensen Huang says
If safety is framed as an engineering problem owned by each vendor, the center of gravity shifts toward technical standards and away from hard law. For operators, that means your actual guardrails will be set by the capabilities and defaults of your suppliers’ stacks more than by regulators for the next few years.
Applied AISalesforce has built the TSA a new AI agent to make travelling less awful for everyone
Government workflows like TSA screening are moving from static rules to AI-mediated triage—once that pattern lands in airports, it will spread to every high-friction public interface. If you sell into public sector or regulated services, assume your buyers will soon expect agentic flows that can justify decisions under audit, not just chatbots at the edge.