Tensions over Taiwan and AI Safety Dialogue Ahead of the September 24 U.S.-China Summit, and South Korea’s Response

Executive Summary

Ahead of the September 24 U.S.-China summit, Beijing has set a red line, stating it could cancel the summit if the U.S. approves new arms sales to Taiwan. In the AI safety dialogue, the conceptual gap is not narrowing, with President Trump’s rejection of calls to slow down AI development clashing with China’s demand to share authority in defining safety. Even if the summit proceeds as planned, fundamental disagreements over the Taiwan issue and the definition of AI safety are highly likely to be carried over unresolved. The diplomatic track of the summit and the regulatory track of agencies like the FCC are already operating separately, making it probable that individual export control measures will continue to accumulate regardless of any symbolic agreements at the summit. The South Korean government and companies should avoid overreacting to the summit's outcome and instead establish a dual-track system that separates security-linked sectors from commercial domains in their response.

I. Analysis of the Current Situation

Maneuvering over Taiwan and AI Safety Dialogue Ahead of the September 24 U.S.-China Summit: An Analysis of the Current Situation

1. Background and Developments

The Trump-Xi summit is a follow-up measure agreed upon during Trump's visit to Beijing last May [9]. President Trump personally announced that Washington plans to honor the visit with an exceptional state dinner [9]. This differs from the 2023 APEC summit in San Francisco, which was aimed at mending relations strained by the spy balloon incident [9]. This time, the autumn visit to the U.S. builds on the preceding diplomatic engagement of the May visit to Beijing, emphasizing continuity in relationship management.

The two countries are projecting a tone of managed competition by restoring high-level military contacts and holding parallel discussions on tariff reductions [9]. However, beneath this superficial stability, the Taiwan issue is solidifying into a structural red line rather than a bargaining chip [9]. As discussions on an arms sales package for Taiwan coincide with moves to increase the U.S. defense budget, Beijing is perceiving these developments as a genuine threat, not merely as negotiating pressure.

The AI safety dialogue is also not a new agenda item. At the May summit in Beijing, the two leaders had already agreed to launch an intergovernmental dialogue on AI [3]. The upcoming Washington summit serves as a follow-up to that agreement [9]. However, the current phase of talks is reaffirming that the two countries had differing political objectives for the term 'safety' from the moment the agreement was made [3].

2. Current Situation

On September 12, the French-language outlet La Nouvelle Tribune, citing The Telegraph, reported that Beijing has set a tangible red line for Washington: the summit itself could be called off if the United States approves new arms sales to Taiwan [10]. The Diplomat also reported that ahead of President Xi Jinping’s visit, Beijing has signaled the possibility of canceling the summit if there are additional arms sales to Taiwan [12]. The White House has not issued an official statement on the arms sales issue, but the U.S. representative in Taiwan recently described securing peace in the Taiwan Strait as one of America’s top priorities [12]. This demonstrates a dual approach from the U.S. side, simultaneously issuing signals of silence and engagement.

On the AI safety dialogue track, the discussion has grown more complex as calls from within the industry to slow down AI development became a public issue just before the summit. During a visit to Ireland on September 13, President Trump met with reporters and dismissed the idea of slowing down AI as a “sick conspiracy,” arguing that such a move would ultimately only benefit China [4]. This stance represents a complete rejection of some industry concerns, framed within the context of maintaining a competitive edge [3]. For its part, China, through commentaries affiliated with state media, has presented preconditions: that the authority to define safety must be shared with the United States, and that the same standards must be applied to American companies [3]. This can be interpreted as an attempt by Beijing to shift the discussion from technological risk to the issue of global rule-making authority [3]. The Spanish newspaper El País reported that China is pushing back, framing the call for a slowdown by Anthropic CEO Dario Amodei as a hidden agenda to stifle its own technological progress [19].

At the Beijing Xiangshan Forum on September 16, Chinese Minister of National Defense Dong Jun called for expanding security dialogues and rule-making in emerging domains such as AI, space, and the deep sea [15]. This can be read as a strategic move by China's military leadership to link technology-related dialogues with the security agenda ahead of the summit.

3. Key Actors and Positions

China (Beijing): Is framing the Taiwan issue as an absolute red line, not a subject for negotiation [10][12]. The remarks by Minister of National Defense Dong Jun at the Xiangshan Forum reaffirmed the position that rule-making authority in new technology domains like AI and space must be shared with the United States [15]. In the AI safety discussions, China is demanding shared authority in defining safety and the application of the same standards to U.S. companies as preconditions, attempting to shift the discourse from technological risk to a matter of leadership in rule-making [3]. Commentaries affiliated with state media portray the U.S. focus on safety as a hidden agenda to contain China’s technological development [19].

United States (White House): President Trump is sending mixed signals by maintaining official silence on arms sales to Taiwan while elevating the status of the summit itself with preparations for a state dinner [9][12]. On AI policy, the administration treats safety solely within the framework of maintaining a competitive edge over China through deregulation, and it squarely rejects industry calls to slow down development [3][4]. The Brookings Institution projects that the headline achievement of this visit will likely be an agreement to extend the trade truce agreed upon in Busan last October. In return, it analyzes that Xi will demand Trump’s continued forbearance on issues related to Taiwan, tariffs, and technology regulations [8].

U.S. Regulatory Agencies and Congress: Separate from the White House's diplomatic track, technology regulations targeting China from agencies like the FCC are accumulating [6]. Given that the flashpoint of this conflict lies within the bureaucratic regulatory track rather than the White House, it is highly likely that individual regulations will continue regardless of whether the summit takes place [6].

AI Industry (U.S.): Key players in frontier AI development, such as Anthropic CEO Dario Amodei, are publicly raising concerns about the pace of technological advancement [19]. Their calls for a slowdown have influenced the summit's agenda but were rejected by the Trump administration within the framework of maintaining a competitive edge [4].

Nvidia: CEO Jensen Huang's expected attendance at the summit suggests that semiconductors and AI hardware have emerged as substantive agenda items for the talks [14].

Taiwan: Despite the renewed threat of U.S. semiconductor tariffs, Taiwan appears relatively unshaken, believing it has already secured preferential treatment through a memorandum of understanding signed with the United States [17].

4. Summary of Key Issues

The first issue is the timing and scale of arms sales to Taiwan. If a U.S. decision to approve a sale conflicts with the red line set by Beijing, the fate of the summit itself becomes uncertain [10][12].

The second issue is the authority to define the concept of AI 'safety.' The U.S. treats safety as a concept subordinate to maintaining its competitive edge, while China frames it as a matter of distributing global rule-making power [3]. As long as this definitional gap remains, any agreement reached at the summit is likely to be merely symbolic [1][3].

The third issue is the separation between the diplomatic track and the regulatory track. Regardless of cooperative signals from the White House, individual measures by regulatory bodies like the FCC are accumulating separately. This means the summit's outcome will not signify a real end to frictions over semiconductors and AI [6].

II. In-Depth Analysis

Maneuvering over Taiwan and AI Safety Dialogue Ahead of the September 24 U.S.-China Summit: An In-Depth Analysis

1. Analysis of Root Causes

At the root of the current situation is the fact that the U.S. and China have fundamentally different understandings of the very grammar of negotiation in an era of 'hegemonic transition.' Washington continues to treat the Taiwan issue as a manageable variable. The Trump administration is seen as having effectively abandoned the principle of strategic ambiguity toward Taiwan [16]. Nevertheless, the White House has not released an official statement on whether it will approve arms sales [12]. This can be read as an intention to continue using the Taiwan card as negotiating leverage that can be traded for the success of the summit.

Beijing's calculus is different. China is redefining the Taiwan issue not as a card on the negotiating table, but as a precondition for negotiations to even take place. The warning that it will cancel the summit if arms sales are approved [10][12] is tantamount to a declaration that it will not treat Taiwan as an issue that can be traded for other agenda items. For President Xi Jinping, the Taiwan issue is directly linked to his domestic political legitimacy. The moment he is seen as trading the Taiwan issue for a diplomatic achievement like a summit, the political cost he would have to bear within Beijing would become greater than that of the arms sale itself.

The deadlock in the AI safety dialogue stems from a more fundamental conceptual clash. President Trump narrowly understands AI safety as an issue of maintaining speed in the technological race. He dismissed calls from some in the industry to slow down as a “sick conspiracy,” arguing that such claims would ultimately only benefit China [4]. Underlying this remark is the concern that the concept of safety could lead to a competitive disadvantage.

China's approach is different in nature. Beijing seeks to reframe the safety discussion from a matter of technological risk management to one of distributing global rule-making authority [3]. This is why it has put forward preconditions, such as sharing the authority to define safety with the U.S. and applying the same standards to American companies [3]. Commentaries affiliated with Chinese state media frame the calls for a slowdown from the U.S. AI industry, including Dario Amodei, as a hidden agenda to stifle China's own technology [19]. This difference in interpretation surrounding the word 'safety' is not a simple communication failure but stems from a fundamental disagreement over who has the authority to define technological standards.

2. Structural Context

Political Structure

The current situation is structured around two different, parallel tracks. One is the summit diplomacy track led by the White House, which includes preparations for the state dinner, the restoration of high-level military contacts, and discussions on tariff reductions [9]. The other is the bureaucratic track led by individual regulatory agencies like the FCC and by Congress. These two tracks are already operating separately [3][6]. Even if a symbolic agreement is reached between the leaders, individual regulations such as export controls on semiconductors and AI are likely to continue under their own momentum [3][6]. Beijing has so far refrained from direct retaliation against measures like those from the FCC, but its patience is now being tested [6].

The configuration of alliances surrounding the Taiwan issue is also part of this structure. Japanese Prime Minister Takaichi's prior communications demonstrate a common response logic among neighboring countries concerned about being excluded from U.S.-China negotiations [9]. The recent statement by the U.S. representative in Taiwan, who called peace in the Taiwan Strait a top priority [12], also suggests that, separate from the White House's official silence, there are moves at the working policy level to quell concerns about Taiwan being abandoned.

Economic Structure

On the economic track, a key agenda item is whether to extend the trade truce agreed upon in Busan last October [8]. The Brookings Institution sees the potential for an agreement to extend this truce to be the headline achievement of the visit [8]. In return, President Xi is likely to demand President Trump's continued restraint on the Taiwan issue, new tariffs, and technology-related sanctions [8]. In other words, the basic framework of this summit is a structure where economic compromise is effectively exchanged for security restraint.

The semiconductor supply chain is also intertwined in this structure. Although President Trump has again mentioned relocating Taiwan's semiconductor manufacturing to the U.S. [17], the Taiwanese side appears largely unfazed, believing it has already secured preferential treatment through a memorandum of understanding with the United States [17]. The expected attendance of Nvidia CEO Jensen Huang at the summit [14] also shows that the semiconductor and AI industries have a tangible stake in the outcome of the diplomatic talks.

Security Structure

In the security domain, competition over rule-making in emerging areas such as AI, space, and the deep sea is emerging as a new axis. At the Xiangshan Forum, Chinese Minister of National Defense Dong Jun called for expanding security dialogues and rule-making in the domains of AI, space, and the deep sea [15]. This shows that U.S.-China competition is extending beyond traditional military rivalry into a struggle for leadership in setting international norms for emerging technologies. If the Taiwan issue is a red line in traditional security, the AI safety dialogue carries the potential for a similar competition to set red lines in the emerging security domain.

3. Historical Precedents and Comparison with Similar Cases

The 2023 APEC summit in San Francisco is a frequently cited precedent for this meeting. That summit was largely aimed at mending relations strained by the spy balloon incident [9]. The upcoming Washington summit is different in nature. As a follow-up visit building on the preceding diplomatic engagement of the May trip to Beijing, it is less about restoring relations and more about testing whether an already established track of cooperation can be maintained [9].

Comparisons to Cold War-era U.S.-Soviet nuclear arms control negotiations are also frequently invoked in discussions of the AI safety dialogue. However, a key difference is that nuclear arms control was based on a shared understanding of a clear physical threat—mutually assured destruction—whereas with AI safety, the two countries have not even agreed on the nature of the threat itself. The U.S. defines AI safety as a technical risk of losing control, while China defines it as a political issue of distributing rule-making authority [3]. This definitional gap is a factor that makes agreement much more difficult to achieve at a more fundamental level than was the case in nuclear arms control negotiations.

The method of 'declaring a red line' on the Taiwan issue is also similar to a pattern repeated in the past regarding the South China Sea. Beijing has a history of presenting clear prohibitive lines on specific issues in advance and then holding the entire relationship hostage if those lines are crossed. Signaling the potential cancellation of the summit over arms sales [10][12] is a continuation of this pattern. However, unlike past cases, the level of pressure is higher this time, as it holds hostage a high-level diplomatic event for which a state dinner has already been prepared.

4. Key Variables Shaping Future Developments

The first variable is the timing and scale of the approval of arms sales to Taiwan. Warnings have already been issued that if the White House approves a new package before the summit, the meeting itself could be canceled [10][12]. Even if the approval is postponed until after the summit, the scale and timing could affect the sustainability of any agreements reached during the talks.

The second variable is the political ripple effect of calls from the AI industry to slow down development. Warnings from industry figures like Dario Amodei have already become a public issue in both countries [2][7][19]. Given that President Trump has publicly rejected these calls [4], the general assessment is that it is unlikely this discussion will lead to a substantive agreement beyond the symbolic level at the summit [1][5]. However, the possibility remains that partial cooperative achievements could be announced in less politically sensitive areas, such as cybersecurity and healthcare [18].

The third variable is the degree of separation between the White House track and the regulatory agency track. Even if a symbolic agreement is reached at the summit, individual measures at the agency level, such as by the FCC, are likely to continue under their own momentum [3][6]. The point at which Beijing's long-held policy of refraining from retaliation crosses a threshold and shifts to active retaliation will determine the volatility of the situation going forward [6].

The fourth variable is whether the trade truce is extended and under what conditions. With the extension of the Busan agreement from last October being discussed as a key outcome of this summit [8], the specific level of restraint Beijing demands in return on Taiwan, tariffs, and technology regulations will determine the actual direction of U.S.-China relations after the summit [8].