Reuters reported on Sept. 4 that U.S. and Chinese officials were preparing for a possible AI-safety dialogue in mid-September, citing two people familiar with the discussions. The report described planning, not a confirmed meeting.
A White House official subsequently told Reuters: “There is currently no planned AI-related meeting in mid-September.” That denial conflicts with the anonymous-source account. The talks should therefore be treated as reported and under discussion, not scheduled.
Reuters sources said the potential agenda could include monitoring AI-driven cyberattacks and information-sharing between labs, as well as concerns about alleged Chinese distillation of U.S. models. Those topics are part of the reported, unverified planning; neither a venue nor participants have been publicly confirmed. Reuters also reported that Treasury Secretary Scott Bessent may lead the U.S. side, another detail not independently confirmed.
The diplomatic stakes are larger than the latest model-launch cycle. Reuters’ sources said China viewed an AI-safety dialogue as a possible deliverable ahead of a summit, suggesting that guardrails and crisis channels could be used to show practical progress even as technology competition continues. That framing remains a source-based account, not an announced agreement.
A Trump-Xi summit in Washington for Sept. 24 has been reported separately. It should not be conflated with a mid-September AI meeting, which the White House says is not currently planned.
Until Washington or Beijing confirms a date, format or participants, the safest description is a possible dialogue reported by Reuters and disputed by the White House—not a confirmed bilateral session.