U.S.–China AI Safety Talks Are Being Discussed, but the Meeting Is Not Yet Confirmed

Planning is reported, but the calendar is disputed

WASHINGTON — American and Chinese officials are discussing possible mid-September talks focused on AI safety, according to Reuters sources briefed on planning. Treasury Secretary Scott Bessent could lead the U.S. side, while China could send Vice Premier He Lifeng or another senior official.

The strongest caveat appears in the same report: a White House official said there is currently no planned AI-related meeting in mid-September. Reuters also said the location, final agenda and participants remain in flux. Those facts require a conditional description.

Diplomatic discussions often begin through informal contacts and staff work before leaders approve a meeting. Such preparation can be significant without guaranteeing the event. Cancellation or postponement would show that planning did not become a final commitment.

Cybersecurity and crisis communication are central

Reuters reported that Washington wants cooperation on monitoring AI-directed cyberattacks and information sharing among laboratories. The focus reflects shared vulnerability: a capable system deployed recklessly or compromised could create cross-border effects faster than ordinary diplomatic channels respond.

The governments remain competitors over chips, models, trade and security. A safety channel would not erase those disputes or certify identical rules. Its narrower value could be establishing whom to call, what evidence to share and how to distinguish an accident from deliberate state action.

NIST's public framework emphasizes identifying, measuring and managing AI risk. International diplomacy adds hard questions about verification and reciprocity. Voluntary promises are useful only when procedures, incentives and transparency make them credible.

A dialogue would not be a treaty

Even if talks occur, an initial exchange is unlikely to create a binding treaty. It could yield a statement, working group or agreement to continue talking. Each outcome has different weight, and coverage should identify what actually happened.

The proposed dialogue comes before a planned September 24 Trump-Xi summit in Washington. That creates an incentive to present a deliverable, but also a risk that symbolic alignment will be mistaken for enforceable safeguards. Published text and implementation matter more than ceremony.

The Treasury Building photograph is an authentic public-domain image selected because Bessent is reported as a possible U.S. lead. It does not depict negotiations, prove that a meeting is booked or show a proposed Chinese participant.

Limits and what to watch

The meeting, location, participants and agenda remain unsettled, and the White House told Reuters no mid-September AI meeting is currently planned. Preparations are not a confirmed event or agreement.

The next development to watch is an official schedule, named delegations, a published agenda or a joint account of completed talks.