Implications of the New U.S.-China AI Emergency Communication Channel and the Introduction of the Term 'Superintelligence'

Executive Summary

The Superintelligence (SI) Dialogue and the AI incident notification channel announced at the U.S.-China summit on September 25 are not products of independent security needs, but rather byproducts of a broader trade package negotiation encompassing tariffs, critical minerals, and Iran-related issues. The introduction of the term "superintelligence" was also pre-established by President Trump as a domestic political discourse on Truth Social prior to his UN General Assembly speech; since China has not yet officially adopted the term, it is difficult to view this as a substantive normative agreement between the two sides. Within Washington, the White House track signaling cooperation and the regulatory track pursuing advanced chip export controls and model distillation investigations operate separately, meaning that symbolic agreements between the leaders do not prevent the continuation of individual regulations against China. Whether notification procedures are fleshed out at the follow-up meeting in Shenzhen in November, along with the readjustment of export controls following the expiration of the Busan Agreement, will serve as key tests of this channel's actual credibility. Rather than aligning its policy priorities with the symbolic announcements of the bilateral U.S.-China channel, South Korea needs to independently manage semiconductor supply chain risks through a dual response system that distinguishes security-linked areas from commercial domains.

I. Issue Analysis

Establishment of the U.S.-China AI Emergency Communication Channel and the Introduction of the Term 'Superintelligence': Issue Analysis

1. Background and Progress of the Issue

This agreement is not an unexpected outcome. At the Beijing summit in May, President Trump and President Xi Jinping had already agreed in principle to initiate an intergovernmental dialogue on AI [9]. The Washington meeting on September 24 serves as a follow-up to this agreement [9]. President Xi's visit carries symbolic weight as his first visit to the White House in 11 years [9]. This symbolism is further underscored by President Trump's personal statement that he had prepared a state dinner [9].

Working-level preparations had been underway prior to the summit. On September 20, U.S. Treasury Secretary Bessent and Chinese Vice Premier He Lifeng met at the JPMorgan Chase headquarters in New York [3]. The meeting was attended by U.S. Trade Representative Greer, China's International Trade Representative Li Chenggang, and Vice Minister of Finance Liao Min [3]. Notably, this discussion did not begin as an independent consultation on AI safety; rather, it was treated as part of a package negotiation encompassing tariffs, critical minerals, and Iran-related issues [3]. The definition and procedures for AI incidents subject to notification remain undecided and have been deferred to the follow-up meeting in Shenzhen in November [3].

The controversy over the terminology shift has its own context. In a Truth Social post on September 19, President Trump characterized the AI safety debate as "an extension of the hoax created by radical left Democrats to destroy America" [6]. In the same post, he conducted a poll of his followers to rename AI, proposing candidates such as "Superior Intelligence" [6]. Three days later, in his speech at the UN General Assembly on September 22, he flatly rejected the concept of an international cooperative body for AI control and declared that official government documents would use the term "superintelligence" instead of AI [6]. This reflects the Trump administration's consistent pattern of responding through security organization rather than regulation [6].

2. Current Situation

In a factsheet released on September 25 (local time) at the conclusion of the summit, the White House specified two key outcomes. One is the establishment of a U.S.-China Superintelligence (SI) Dialogue to "exchange views on the risks and benefits of superintelligence" [1], with the next meeting scheduled by November [1]. The other is a bilateral communication channel for mutual notification in the event of AI incidents [16]. The two countries agreed at this month's summit to use the term "superintelligence" in the future when referring to new technologies related to "artificial intelligence" [16].

There is a clear divergence in temperature regarding the adoption of the terminology. The Hankyoreh pointed out that "superintelligence" is interpreted as a "political expression" by President Trump, emphasizing development over technology control, and reported that China has not yet officially used this expression [1]. In other words, the terminology shift is closer to framing for U.S. domestic politics, and there is no verified evidence that China has substantively accepted it.

Around the same time, President Xi remarked that AI must not escape human control, while President Trump stated that they had agreed that new regulations are unnecessary [9]. This appears to be a gesture for domestic politics on both sides rather than a normatively binding achievement [9]. Within Washington, the White House track emphasizing cooperation and the regulatory track conducting advanced chip export controls and model distillation investigations operate separately [3]. Given the precedent of the 2024 U.S.-China AI dialogue collapsing after becoming entangled with trade and Taiwan issues, this agreement is highly likely to remain a symbolic item to fill the summit's list of achievements rather than functioning as substantive crisis management infrastructure [3]. In November, when the U.S.-China Busan Agreement expires, whether export controls are tightened again will serve as the first test of this channel's credibility [3].

3. Key Actors and Positions

The White House / President Trump: President Trump's priority is maintaining the pace of AI development. The introduction of the term "superintelligence" and the rejection of international regulatory frameworks are read as symbolic measures supporting this priority [6]. The Treasury Department (Secretary Bessent) puts forward a pragmatic frame of securing transparency toward China. Secretary Bessent noted, "Just like other cross-border activities, moving from opacity to transparency between the world's first and second AI powers is extremely important" [12]. However, this statement reflects the perspective of the Treasury track, which emphasizes the need for cooperation, and differs in tone from the tracks aimed at countering China, such as the Commerce Department and Congress [3][9].

The Chinese Side (President Xi Jinping, Vice Premier He Lifeng, Ministry of Commerce Line): China exhibits an attitude of utilizing safety discussions as a means to secure global rule-making authority [9]. President Xi's "human control" remark can be interpreted as expressing principled agreement on AI risks while signaling that China will not be subordinated to a U.S.-style regulatory frame [9]. Its non-adoption of the term "superintelligence" is also consistent with this distancing [1].

U.S. Congress and Regulatory Agencies: Semiconductor export controls and model distillation investigations continue regardless of the White House's cooperative signaling [3][6]. This demonstrates that symbolic agreements between leaders do not prevent the continuation of individual regulations [9].

The AI Industry (U.S. Frontier Labs): In the same context, it is noteworthy that the heads of competing U.S. AI companies have recently made an unusual joint call for "restraint" [8]. This reflects a sense of crisis within the industry regarding the need to manage the risks of superintelligence-level models.

International Think Tanks and Expert Groups: Organizations such as the CFR, Brookings, and the Carnegie Endowment for International Peace have consistently raised the need for U.S.-China AI arms control negotiations [2][5][11]. The Brookings-Tsinghua Track II dialogue, which has discussed principles of human control over military AI since 2019, serves as a precedent and can be seen as part of the conceptual background for this summit-level agreement [13].

4. Key Issues

The first issue is the effectiveness of the communication channel. The definition and notification procedures for AI incidents have not yet been finalized [3]. There is a precedent of the 2024 U.S.-China AI dialogue collapsing after becoming entangled with trade and Taiwan issues [3]. The circumstances of this agreement being negotiated as a package with tariffs, critical minerals, and Iran-related issues raise questions about the channel's independence [3].

The second issue is the implications of the terminology shift. Naming it "superintelligence" is highly likely a domestic political framing by the Trump administration rather than a substantive risk management measure [1][6]. The fact that China has not yet officially adopted this term suggests that the two countries do not share the same conceptual framework [1].

The third issue is whether the dual-track structure will persist. The coexistence of the White House-led cooperative track and the export control and congressional-led pressure track is not resolved by this agreement [3][6][9]. Whether export controls are tightened again upon the expiration of the Busan Agreement in November will serve as a test to gauge the actual direction of this dual structure [3].

The fourth issue is the divergence in the goals that the U.S. and China assign to safety discussions. While China approaches safety discourse as a means to secure global rule-making authority, the U.S. addresses safety solely within the framework of maintaining its competitive advantage over China [9]. This fundamental divergence in goals is a variable that will determine how substantively the SI Dialogue and the incident notification channel will operate in the future.

II. In-Depth Analysis

Establishment of the U.S.-China AI Emergency Communication Channel and the Introduction of the Term 'Superintelligence': In-Depth Analysis

1. Root Cause Analysis

The ostensible justification for this agreement is AI safety management. However, tracing the actual path of its development reveals a different picture. Discussions on establishing the channel began at the meeting between U.S. Treasury Secretary Bessent and Chinese Vice Premier He Lifeng at the JPMorgan Chase headquarters in New York on September 20 [3]. The agenda for this meeting was a package negotiation encompassing tariffs, critical minerals, and Iran-related issues [3]. AI safety was merely inserted as one item within that negotiation package. In other words, the AI crisis management channel did not originate from independent security needs, but was created as a byproduct of trade and resource negotiations.

The developmental path of the terminology shift is even simpler. On September 19, President Trump defined the AI safety debate itself on Truth Social as "an extension of the hoax created by radical left Democrats to destroy America" [6]. In the same post, he gathered candidate names for a new term through a follower poll [6]. Three days later, in his speech at the UN General Assembly, he rejected the concept of an international AI regulatory cooperative body and formalized the adoption of the term "superintelligence" [6]. Looking at this sequence, the term "superintelligence" was not a product of international negotiations, but was first established in U.S. domestic political discourse and subsequently appended to the summit outcomes. The Hankyoreh also pointed out that this expression is a "political expression" by President Trump, emphasizing development over technology control, and that the Chinese side has not yet officially used it [1]. The root cause is not a shared U.S.-China perception of AI risks, but rather the intersection of the U.S. domestic political landscape and the need to manage the negotiation package.

2. Structural Context

Three levels of structure define this issue.

In the political structure, the White House track and the regulatory/congressional tracks operate separately. EAI analysis points out that "within Washington, the White House track emphasizing cooperation and the regulatory track conducting advanced chip export controls and model distillation investigations operate separately" [3]. This means that symbolic agreements between leaders do not prevent the continuation of individual regulations such as semiconductor export controls [9]. This dual structure is a pattern that has been repeated since the beginning of the Trump administration, where cooperative gestures are projected externally while working-level regulatory agencies maintain a stance of pressure against China.

In the economic structure, the fact that the AI channel was negotiated as a package with tariffs, critical minerals, and Iran-related issues is decisive [3]. This demonstrates that AI safety cooperation is treated not as an independent agenda item, but as a bargaining chip in a broader U.S.-China trade package. The analysis suggesting that the upcoming decision on whether to retighten export controls ahead of the expiration of the Busan Agreement in November will serve as a test of the channel's credibility falls within the same context [3].

In the security structure, the goals that China and the U.S. assign to safety discussions are fundamentally different. EAI analysis assesses that "while China seeks to utilize safety discussions as a means to secure global rule-making authority, the U.S. addresses safety solely within the frame of maintaining its competitive advantage over China, meaning that the goals pursued by the two sides are fundamentally different" [9]. The observation that President Xi remarked at the Washington meeting that AI must not escape human control, while President Trump agreed that new regulations are unnecessary [9], gains credibility as being closer to gestures for domestic politics on both sides rather than a binding normative agreement. This structural misalignment of goals is a fundamental condition that makes it difficult for the channel to develop into substantive crisis management infrastructure.

3. Comparison with Historical Precedents and Similar Cases

The most direct precedent is the 2024 U.S.-China AI dialogue. This dialogue collapsed after becoming entangled with trade and Taiwan issues [3]. The current agreement shares the same structure, as it was also handled as a package alongside tariffs, critical minerals, and Iran-related issues [3]. In light of this precedent, EAI analysis judges that "this agreement is highly likely to remain a symbolic item to fill the summit's list of achievements rather than functioning as substantive crisis management infrastructure" [3].

The Track II dialogue operated by the Brookings Institution and Tsinghua University since 2019 is another reference precedent [13]. This dialogue has addressed military AI risks at the level of private experts rather than through official intergovernmental channels. Although it has persisted for a long period, it has failed to lead to official intergovernmental agreements. This reaffirms the pattern where U.S.-China AI security dialogue is sustainable at lower levels, but becomes subordinated to other trade issues the moment it is elevated to a summit-level agreement.

Comparisons with the Cold War U.S.-Soviet nuclear hotline are frequently invoked in the media, but the structural differences are significant. The nuclear hotline was established with specifically agreed-upon notification targets and procedures based on a clear balance of terror known as mutually assured destruction. In contrast, the current AI channel has deferred the very definition and procedures of AI incidents subject to notification to the follow-up meeting in Shenzhen in November [3]. In other words, although labeled a hotline, it has initiated without any agreement on what to report.

4. Key Variables Shaping Future Developments

The first variable is the expiration of the Busan Agreement in November and whether export controls will be tightened again. EAI defines this moment as "the first test of this channel's credibility" [3]. If export controls are tightened again even while the cooperative channel is operational, the gap between the channel's symbolism and the substantive stance of pressure against China will be exposed.

The second variable is whether the targets and procedures for AI incident notification are actually fleshed out at the follow-up meeting to be held in Shenzhen in November [3]. If a state where only the channel exists without definitions and procedures persists, it is highly likely to repeat the path of collapse seen in the 2024 dialogue.

The third variable is whether China accepts the term "superintelligence." To date, China has not officially used this expression [1]. If China continues to adhere to the existing term "AI," the agreement between the two countries will remain in a state of misalignment starting from the terminology in official documents. This could serve as a touchstone to gauge the level of mutual trust prior to the actual operation of the channel.

The fourth variable is the policy alignment between the White House track and the congressional/regulatory tracks [3]. If summit achievements and working-level regulatory measures continuously diverge, the channel is unlikely to function as anything more than a performance indicator for domestic politics in both the U.S. and China.

From South Korea's perspective, these four variables do not operate in isolation. In particular, EAI analysis repeatedly emphasizes that government departments responsible for semiconductor supply chain management must track the symbolic announcements of the channel separately from working-level export control measures [3][9].