Proposed US-China AI hotline aims to prevent catastrophic mistrust

A recent LessWrong essay by Troy Tian draws direct analogies between the Cuban Missile Crisis and the current US-China AI race, warning that without deliberate trust-building, a miscalculation could be catastrophic. Tian highlights the tragic case of Jane Ying Wu, a Chinese-born American neuroscientist whose lab was closed by the NIH after an uncharged espionage investigation, leading to her suicide in 2024. A 2023 PNAS study found that 72% of Chinese American scientists feel unsafe, and 65% worry about collaboration with China. This environment of suspicion, Tian argues, makes it nearly impossible to recruit the best experts for joint AI safety efforts, mirroring the conditions that could lead to an accidental conflict.

Tian's proposed intervention is a Beijing-Washington hotline dedicated to AI risk assessment, modeled on the direct communication channels that helped de-escalate Cold War tensions. The hotline would be staffed by American and Chinese academics, creating a controlled but transparent environment for sharing research on AI safety evaluations and taxonomies—rather than on capabilities like model weights. The biggest obstacle is the chilling effect on Chinese-American scientists, who fear investigation or guilt by association. Tian suggests that institutionalizing transparent, regulated collaboration (e.g., formal safety assessments) can provide legal and professional protection, gradually rebuilding trust. He calls on the AI governance community to prioritize humanizing scientists on both sides, lest we repeat the mistakes of the past.