The United States and China are preparing for another round of government-to-government talks on artificial intelligence.
The timing is not settled.
Reuters reported on September 4, citing two sources briefed on the planning, that a dialogue focused on AI safety was being prepared for mid-September. The same report said a U.S. Treasury spokesperson indicated the two sides may instead meet in October.
The final agenda, participants, and location also remained unresolved in the reporting.
That distinction matters.
The evidence supports saying that talks are being prepared. It does not yet support saying that a mid-September meeting is officially scheduled.
The channel itself is not speculative
The broader government-to-government dialogue predates the September scheduling report.
On May 19, the Chinese government said President Donald Trump and President Xi Jinping had held constructive exchanges on artificial intelligence and agreed that their governments would hold a dialogue on the subject.
Reuters separately reported the Chinese Foreign Ministry statement at the time.
That gives the underlying channel firmer footing than the current calendar.
The governments have agreed to talk about AI. What remains unsettled is when the next round will occur and exactly what form it will take.
Reuters's September 4 report adds a more specific development: sources briefed on the planning described talks focused on AI safety risks. The report itself also contains evidence that the timetable is still moving.
For the Record, "being prepared" and "scheduled" are different claims.
Only the first is presently supported.
This would not be the first U.S.-China discussion of AI risk
There is an important historical boundary around the story.
The United States and China held an official intergovernmental AI dialogue in Geneva on May 14, 2024. Public records from both governments confirm that the meeting included discussion of risks associated with advanced artificial intelligence.
That means an unqualified description of the proposed 2026 talks as the first U.S.-China AI-safety dialogue would be wrong.
Reuters uses a narrower distinction: if the new talks occur as described, they would be the first official bilateral discussions devoted exclusively to AI since Trump began his second term.
That is a smaller claim than "first ever," and it is the one the current evidence supports when attributed to Reuters.
What is changing
The significant development is not a treaty, a joint safety standard, or a common regulatory system.
None of those exists here.
The change is diplomatic.
Artificial intelligence is becoming a recurring subject of direct bilateral diplomacy between the United States and China.
A dialogue channel does not imply agreement. It does not mean either government has accepted the other's approach to AI governance. And it does not guarantee that a planned meeting will happen on the timetable now being discussed.
The May agreement gives the two governments a bilateral channel they can return to if both sides continue it.
The 2024 Geneva meeting established a precedent. The leaders' May 2026 agreement renewed the commitment to government-to-government AI dialogue. The September planning report says the next round is being organized around AI safety risks.
Whether the channel becomes sustained or produces substantive agreements cannot yet be known.
What remains unknown
As of September 4, the date of the next talks is unresolved.
Mid-September is a reported planning window, not an official date.
October remains possible according to the Treasury response reported by Reuters.
The final agenda, complete participant list, and location have not been publicly established.
There is also no outcome to evaluate because the reported meeting has not happened.
Those unknowns are not reasons to erase the development from the historical record.
They are part of the development.
Why this enters the Record
Diplomatic channels often become visible before their consequences do.
The relevant fact today is that the United States and China have an agreed intergovernmental AI channel and are reported to be preparing another round focused on AI safety while negotiating its timing and form.
That may eventually produce little more than another meeting.
It may also give the two governments a recurring venue for communication about increasingly capable AI systems if both sides continue the channel.
The evidence available today cannot settle that question.
The Record therefore preserves the narrower claim: a bilateral AI dialogue is taking shape, another round is being prepared, and its calendar is still uncertain.
If the timetable changes, the Record should preserve that change as a subsequent development.