On September 22, 2026, US and Chinese officials wrapped up two days of talks in New York focused on artificial intelligence, trade and investment ahead of Xi Jinping’s visit to Washington, according to Bloomberg reporting carried by Moneycontrol. In a separate interview with the South China Morning Post, US Congressman Ro Khanna urged nuclear style safeguards for AI, including a ban on self improving systems and international inspections of top labs.
This article aggregates reporting from 2 news sources. The TL;DR is AI-generated from original reporting. Race to AGI's analysis provides editorial context on implications for AGI development.
These New York talks and Khanna’s call for “nuclear style” AI safeguards show that AI risk is no longer a side note in US China economic diplomacy. When trade, investment and AI governance are negotiated in the same room, compute exports, data flows and model deployment norms become bargaining chips alongside tariffs and market access. For frontier labs, that means the strategic envelope of what they can build and where they can sell it will increasingly be shaped by great power compromises, not just domestic regulators.
Khanna’s specific proposals, like banning self improving AI systems and instituting inspections of major labs, are politically ambitious, but they echo language already circulating in safety circles about recursive self improvement and verification regimes. Even if they are not adopted wholesale, some version of incident reporting, third party evaluations and on site checks for the largest training runs could emerge from this process. China’s counter moves, seen in Huawei’s accelerated chip roadmaps and its own governance initiatives, reinforce that both sides are trying to lock in technical self sufficiency before conceding too much on rules.

