The US and China set up an AI crisis hotline

Assessment

The question appears already satisfied, with a qualifying publication date of September 26, 2026. My 99% near-term probability reflects direct confirmation by both governments, not a prediction that negotiations will soon succeed.

The relevant reference class is summit-announced US–China sector-specific incident-communication agreements, rather than operational hotlines. A close precedent is the September 25, 2015 White House fact sheet, which separately announced a cybercrime dialogue and an agreement to establish a hotline for escalating problems with assistance requests. This is too small and selected a reference class to provide a defensible empirical quarterly success rate. For the residual scenario in which the latest announcement is insufficient, I use a judgmental baseline hazard of 10% per quarter for a qualifying follow-up publication, increased to 25% in Q4 2026 because a November dialogue is expressly scheduled. Direct current evidence overwhelmingly dominates that baseline. Searches did not identify a sufficiently close prediction market to provide an independent price anchor.

Both required official confirmations are public

The White House fact sheet, dated September 25, 2026, states: “The United States and China also agreed to establish a bilateral communication channel for SI incidents.” It separately establishes a Super Intelligence Dialogue, with its next exchange by November. I checked the page’s HTML: its displayed publication date is September 25, while its UTC publication timestamp is September 26. Either is within the resolution window.

China’s Ministry of Foreign Affairs announcement of September 26, 2026, item seven, independently states: “双方同意建立人工智能事件的沟通渠道”—“Both sides agreed to establish a communication channel for artificial-intelligence incidents.” It too separately announces the AI dialogue and November meeting. Thus these are parallel official confirmations of a definite agreement, not a unilateral assertion or an agreement merely to explore the idea. The difference between “SI” and “AI” is not evidence of different mechanisms: the White House explicitly treats SI as its preferred terminology for the emerging technologies, and a September 26 Chinese Foreign Ministry response, reproduced by its embassy, acknowledges and respects the US terminology. The later official publication establishes September 26 as the resolution date under A3. Neither the excluded May summit commitments nor subsequent operational readiness is needed.

Strongest contrary case and residual uncertainty

The strongest objection is textual: neither short announcement spells out urgency, emergency staffing, notification deadlines, or incident thresholds. An unusually strict reading could regard “communication channel for incidents” as insufficiently explicit about urgent response. My adopted reading is that the separately established bilateral incident channel, contrasted with the periodic policy dialogue, designates an incident-response function and qualifies without detailed operating procedures. This is also consistent with the September 26 AP report carried by PBS, although that report is corroboration, not a resolving source. I retain 1% near-term uncertainty for this interpretation risk. If further specificity is required, November’s dialogue and subsequent bilateral meetings provide opportunities to publish it. Conversely, stalled implementation, security sensitivities, or statements remaining vague could prevent qualification under that stricter reading. Actual non-use or later deterioration in relations would not reverse an already qualifying announcement.

Probability construction

I calculated all 21 cumulative probabilities in Python using 1 − ∏(1 − h): an effective first-period hazard of 0.99, reflecting the already-published evidence; then 0.25 for Q4 2026 conditional on non-resolution; and 0.10 for each subsequent quarter. These are judgmental publication hazards, not measured historical frequencies. The resulting cumulative probability rises from 0.9900 to 0.998987. Nearly all probability mass is on the existing September announcement qualifying, rather than on a future agreement.

Question. By the last day of each quarter, have the US and Chinese governments announced a dedicated hotline or crisis-communication mechanism for incidents involving AI systems?

Definitions.

• Governments: the US federal executive branch and the central government of the People's Republic of China, including their ministries, agencies, and armed forces.
• Dedicated hotline or crisis-communication mechanism: a standing government-to-government channel or agreed procedure expressly designated for urgent notification, clarification, consultation, or coordinated response concerning AI-related incidents. Telephone, secure messaging, designated emergency points of contact, and emergency consultation procedures all count. A new AI-specific procedure using an existing diplomatic or military channel counts; merely observing that an existing general-purpose channel could be used does not. The mechanism need not serve AI incidents exclusively, but must have an expressly designated AI-incident function.
• Incidents involving AI systems: actual or suspected events in which the operation, malfunction, loss of control, or misuse of an AI system causes or threatens harm, including civilian, military, and cross-border incidents. No particular model developer, severity threshold, or actual incident is required.
• Announced: both governments publicly confirm that they have established, or agreed to establish, the same qualifying mechanism. A joint statement issued on behalf of both governments is sufficient; otherwise parallel official statements are required. Operational readiness, a published telephone number, and demonstrated use are not required.

Resolves YES if a joint statement or parallel official statements satisfy all definitions within the Resolution window and on or before the quarter-end. A definite agreement to establish the mechanism counts even if implementation is scheduled for later. For parallel statements, the qualifying publication date is the date of the later statement, per A3.

Exclusions (resolve NO).

• Proposals by experts, think tanks, companies, or former officials, without qualifying announcements by both governments.
• Unilateral offers or statements by only one government, including unilateral claims that the other government agreed.
• Agreements merely to discuss, consider, explore, or negotiate a mechanism.
• General AI dialogues, periodic policy meetings, research cooperation, routine incident-data exchanges, or commitments to AI safety that do not establish an urgent incident-communication function.
• General diplomatic or military hotlines without an expressly designated AI-incident function, and one-off communication about an incident without a standing arrangement.

Resolving source. Official joint statements, agreements, government readouts, or official briefing transcripts published by the White House, US Department of State, US Department of Defense, or another responsible US federal executive agency; and by China's Ministry of Foreign Affairs, State Council, Ministry of National Defense, or another responsible central-government agency. Official embassy publications count. Xinhua counts only when carrying the text of an official government statement or agreement, not independent reporting or commentary. These are narrower resolving sources than Public confirmation (A2); media reports and silence under the A2 dispute rule cannot substitute for either government's confirmation. Apply the Resolution window and Resolution date (A3) conventions. The May 2026 general communication commitments described in the official summit readout do not themselves qualify.