The first plank is keeping frontier models away from non-state actors such as criminal groups, cyber operators or private labs seeking dangerous capabilities.
The second is shared best practice: dangerous-capability evaluations, weight security and deployment governance that could become a bilateral safety baseline.
The talks sit beside chip diplomacy. Export controls alone have not stopped China’s AI progress, so Washington is also trying to move from blocking capacity to writing rules.
参考資料: U.S. China AI talks: Bessent says U.S. leads, safety protocol planned（CNBC）, US, China are discussing AI guardrails to safeguard most powerful models, Bessent says（Reuters）, Bessent Says US, China to Launch AI Safety Talks After Trump-Xi Meeting in Beijing（The Epoch Times）; CocoLoop